Medical data processing device and medical data processing method

Abstract

The technical problem is to improve the accuracy of machine learning. A medical data processing device comprises: an acquisition unit that acquires medical data related to a subject; and an output unit that performs a predetermined process on the medical data and outputs medical diagnosis data, and outputs standardized medical data based on the medical data that is standardized for machine learning without performing a part or all of the predetermined process on the medical data.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based upon and claims the benefit of priority of Japanese Patent Application No. 2019-113447, filed on June 19, 2019, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The invention relates to a medical data processing device and a medical data processing method. Background Art

[0004] Machine learning using medical data such as medical images is performed. When machine learning is performed based on medical data collected by a single medical image diagnostic device, a dedicated protocol is used to unify the conditions of the medical data used in the machine learning. However, when machine learning is performed based on medical data collected by various medical image diagnostic devices, differences in the setting facilities or device modifications may cause fluctuations in the medical data. For example, in the same facility, the various collection conditions of medical data are not completely random, and there are many identical settings, so statistical deviations are likely to occur. Due to such fluctuations in medical data, the accuracy of machine learning is reduced.

[0005] Prior art literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2019-10410

[0007] Patent Document 2: Japanese Patent Application No. 2019-502418

[0008] Patent Document 3: Japanese Patent Application Publication No. 2006-301965

[0009] Patent Document 4: International Publication No. 2007 / 029467 Summary of the invention

[0010] The problem to be solved by the present invention is to improve the accuracy of machine learning.

[0011] The medical data processing device of the embodiment comprises: an acquisition unit, which acquires medical data related to a subject; and an output unit, which performs prescribed processing on the medical data and outputs medical diagnostic data, and outputs standardized medical data based on the medical data that is standardized for machine learning without performing part or all of the prescribed processing.

[0012] Effects of the Invention

[0013] Can improve the accuracy of machine learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a diagram showing the configuration of a magnetic resonance imaging apparatus according to the first embodiment.

[0015] Figure 2 This is a diagram showing the flow of data output processing by the processing circuit.

[0016] Figure 3 This is a diagram showing the flow of other data output processing by the processing circuit.

[0017] Figure 4 This is a diagram showing the flow of other data output processing by the processing circuit.

[0018] Figure 5 This is a diagram showing the flow of other data output processing by the processing circuit.

[0019] Figure 6 FIG. 1 is a diagram schematically showing output formats of medical diagnosis data and standardized medical data.

[0020] Figure 7 This is a diagram showing an example of a GUI screen related to data output processing.

[0021] Figure 8 It is a data set that displays medical diagnostic data and standardized medical data with annotations. Figure 7 FIG. 1 is a diagram showing an example of a GUI screen of .

[0022] Fig. 9 It is a diagram showing the configuration of an X-ray computed tomography apparatus according to a second embodiment.

[0023] Fig.10 It is a diagram showing the flow of data output processing by the processing circuit according to the second embodiment.

[0024] Fig.11 It is a diagram showing the configuration of an ultrasonic diagnostic apparatus according to a third embodiment.

[0025] Fig.12 It is a diagram showing the flow of data output processing by the processing circuit according to the third embodiment.

[0026] Description of Reference Numerals

[0027] 51 Processing circuit

[0028] 52 Memory

[0029] 53 Display

[0030] 54 Input Interface

[0031] 55 Communication interface

[0032] 511 Get Function

[0033] 512 Output function

[0034] 513 Display control function DETAILED DESCRIPTION

[0035] A medical data processing device according to one embodiment includes an acquisition unit and an output unit. The acquisition unit acquires medical data related to a subject. The output unit performs a predetermined process on the medical data and outputs medical diagnosis data, and does not perform a part or all of the predetermined process and outputs standardized medical data based on the medical data that is standardized for machine learning.

[0036] The medical data processing device of this embodiment is a computer or processor that processes medical data collected by a medical device. As a medical device of this embodiment, a medical image diagnostic device and a biological information measuring device can be used. The medical image diagnostic device collects medical images by performing medical imaging on the subject through various imaging principles. As medical image diagnostic devices, for example, there are magnetic resonance imaging devices, X-ray computed tomography devices, ultrasonic diagnostic devices, nuclear medicine diagnostic devices, X-ray diagnostic devices, optical interference tomography devices, optical ultrasonic devices, endoscopes, etc. The biological information measuring device collects waveform data related to the biological information of the subject through various measurement principles. As biological information measuring devices, for example, there are automatic analyzers, electrocardiographs, respirometers, sphygmomanometers, pulse oximeters, etc.

[0037] The medical data processing device of this embodiment may be a computer or processor mounted on the medical device, or a computer or processor separate from the medical device. For specific description, the medical data processing device of the first embodiment is a computer mounted on a magnetic resonance imaging device, the medical data processing device of the second embodiment is a computer mounted on an X-ray computed tomography device, and the medical data processing device of the third embodiment is a computer mounted on an ultrasonic diagnostic device.

[0038] (First Embodiment)

[0039] Figure 1 1 is a diagram showing the structure of a magnetic resonance imaging apparatus 1 according to the first embodiment. Figure 1 As shown, the magnetic resonance imaging apparatus 1 includes a gantry 10 , a bed 30 , a gradient magnetic field power supply 21 , a transmitting circuit 23 , a receiving circuit 25 , a bed driving device 27 , a sequence control circuit 29 , and a medical data processing device 50 .

[0040] The gantry 10 has a static magnetic field magnet 41 and a gradient magnetic field coil 43. The static magnetic field magnet 41 and the gradient magnetic field coil 43 are accommodated in the housing of the gantry 10. A hollow hole is formed in the housing of the gantry 10. A transmitting coil 45 and a receiving coil 47 are arranged in the hole of the gantry 10.

[0041] The static magnetic field magnet 41 has a hollow cylindrical shape and generates a static magnetic field inside the cylindrical shape. As the static magnetic field magnet 41, for example, a permanent magnet, a superconducting magnet, or a normal conductive magnet is used. Here, the central axis of the static magnetic field magnet 41 is defined as the Z axis, the axis perpendicular to the Z axis is defined as the Y axis, and the axis perpendicular to the Z axis is defined as the X axis. The X axis, the Y axis, and the Z axis constitute an orthogonal three-dimensional coordinate system.

[0042] The gradient magnetic field coil 43 is a coil unit installed on the inner side of the static magnetic field magnet 41 and formed into a hollow, roughly cylindrical shape. The gradient magnetic field coil 43 receives the supply of current from the gradient magnetic field power supply 21 to generate a gradient magnetic field. In more detail, the gradient magnetic field coil 43 has three coils corresponding to the mutually orthogonal X-axis, Y-axis, and Z-axis. The three coils form a gradient magnetic field in which the magnetic field strength varies along the X-axis, Y-axis, and Z-axis. The gradient magnetic fields along the X-axis, Y-axis, and Z-axis are synthesized to form mutually orthogonal slice selection gradient magnetic fields Gs, phase encoding gradient magnetic fields Gp, and frequency encoding gradient magnetic fields Gr in the desired direction. The slice selection gradient magnetic field Gs is used to arbitrarily determine the imaging section (slice). The phase encoding gradient magnetic field Gp is ​​used to change the phase of the magnetic resonance signal (hereinafter referred to as the MR signal) according to the spatial position. The frequency encoding gradient magnetic field Gr is used to change the frequency of the MR signal according to the spatial position. In the following description, it is assumed that the slice selection gradient magnetic field Gs is tilted in the Z-axis direction, the phase encoding gradient magnetic field Gp is ​​tilted in the Y-axis direction, and the frequency encoding gradient magnetic field Gr is tilted in the X-axis direction.

[0043] The gradient magnetic field power supply 21 supplies current to the gradient magnetic field coil 43 according to the sequence control signal from the sequence control circuit 29. The gradient magnetic field power supply 21 supplies current to the gradient magnetic field coil 43, so that the gradient magnetic field coil 43 generates a gradient magnetic field along each axis of the X axis, the Y axis, and the Z axis. The gradient magnetic field is superimposed on the static magnetic field formed by the static magnetic field magnet 41 and applied to the subject P.

[0044] The transmission coil 45 is disposed, for example, inside the gradient magnetic field coil 43 , and generates a high-frequency magnetic field pulse (hereinafter referred to as an RF magnetic field pulse) by receiving a supply of current from the transmission circuit 23 .

[0045] The transmission circuit 23 supplies current to the transmission coil 45 in order to apply an RF magnetic field pulse for exciting the target protons present in the subject P to the subject P via the transmission coil 45. The RF magnetic field pulse vibrates at a resonance frequency inherent to the target protons to excite the target protons. MR signals are generated from the excited target protons, and the MR signals are detected by the receiving coil 47. The transmission coil 45 is, for example, a whole body coil (WB coil). The whole body coil can also be used as a transceiver coil.

[0046] The receiving coil 47 receives MR signals emitted from target protons present in the subject P under the action of the RF magnetic field pulse. The receiving coil 47 has a plurality of receiving coil elements capable of receiving MR signals. The received MR signals are supplied to the receiving circuit 25 via wire or wirelessly. Figure 1 Although not shown in the figure, the receiving coil 47 has a plurality of receiving channels installed in parallel. The receiving channel has a receiving coil element for receiving MR signals and an amplifier for amplifying the MR signals. MR signals are outputted for each receiving channel. The total number of receiving channels may be the same as the total number of receiving coil elements, and the total number of receiving channels may be more or less than the total number of receiving coil elements.

[0047] The receiving circuit 25 receives MR signals generated from the excited object protons via the receiving coil 47. The receiving circuit 25 performs signal processing on the received MR signals to generate digital MR signals. The digital MR signals can be represented by the k-space defined by the spatial frequency. Therefore, hereinafter, the digital MR signals are referred to as k-space data. K-space data is a type of raw data for image reconstruction. The k-space data is supplied to the medical data processing device 50 via wired or wireless communication.

[0048] The above-mentioned transmission coil 45 and reception coil 47 are merely examples. A transmission coil having a transmission function and a reception function may be used instead of the transmission coil 45 and the reception coil 47. In addition, the transmission coil 45, the reception coil 47, and the transmission coil may be combined.

[0049] A diagnostic bed 30 is provided adjacent to the gantry 10. The diagnostic bed 30 includes a top plate 33 and a base 31. A subject P is placed on the top plate 33. The base 31 supports the top plate 33 so as to be able to slide along the X-axis, the Y-axis, and the Z-axis, respectively. A diagnostic bed drive device 27 is accommodated in the base 31. The diagnostic bed drive device 27 receives control from a sequence control circuit 29 to move the top plate 33. The diagnostic bed drive device 27 may include any motor such as a servo motor or a stepping motor.

[0050] The sequence control circuit 29 has a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), and a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory) as hardware resources. The sequence control circuit 29 synchronously controls the gradient magnetic field power supply 21, the transmission circuit 23, and the reception circuit 25 based on the imaging protocol determined by the processing circuit 51, etc., performs MR imaging on the subject P according to the pulse sequence corresponding to the imaging protocol, and collects raw data related to the subject P.

[0051] like Figure 1 As shown, the medical data processing device 50 is a computer having a processing circuit 51 , a memory 52 , a display 53 , an input interface 54 , and a communication interface 55 .

[0052] The processing circuit 51 has a processor such as a CPU as a hardware resource. The processing circuit 51 functions as a core of the magnetic resonance imaging apparatus 1. For example, the processing circuit 51 has an acquisition function 511, an output function 512, and a display control function 513 by executing various programs.

[0053] In the acquisition function 511 , the processing circuit 51 acquires medical data related to the subject P. The acquired medical data is, for example, raw data collected via the receiving circuit 25 , in other words, k-space data.

[0054] In the output function 512, the processing circuit 51 performs a predetermined process on the raw data and outputs medical diagnosis data, and does not perform a part or all of the predetermined process and outputs standardized medical data based on the raw data that is standardized for machine learning. Medical diagnosis data is data generated based on raw data and is data for medical diagnosis. Standardized medical data is medical data for machine learning and is standardized for machine learning. For example, the standardized medical data is used as input data used in parameter learning of a machine learning model.

[0055] In the display control function 513 , the processing circuit 51 displays various information on the display 53 . For example, the processing circuit 51 displays the medical diagnosis data and the standardized medical data generated by the output function 512 on the display 53 .

[0056] The memory 52 is a storage device such as a HDD (Hard Disk Drive), an SSD (Solid State Drive), or an integrated circuit storage device that stores various information. In addition, the memory 52 may also be a drive device that reads and writes various information between a removable storage medium such as a CD-ROM drive or a DVD drive, or a flash memory. For example, the memory 52 stores raw data, k-space data, medical diagnosis data, standardized medical data, and various programs.

[0057] The display 53 displays various information through the display control function 513. For example, the display 53 displays medical diagnosis data and standardized medical data output by the output function 512. As the display 53, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the technical field can be appropriately used.

[0058] The input interface 54 includes an input device for receiving various instructions from the user. As the input device, a keyboard, a mouse, various switches, a touch screen, a touch pad, etc. can be used. In addition, the input device is not limited to devices with physical operating parts such as a mouse and a keyboard. For example, a processing circuit for receiving an electrical signal corresponding to an input operation from an external input device provided separately from the magnetic resonance imaging apparatus 1 and outputting the received electrical signal to various circuits is also included in the example of the input interface 54.

[0059] The communication interface 55 is an interface for connecting the magnetic resonance imaging apparatus 1 to a computer for machine learning. The computer for machine learning is, for example, a computer included in a PACS (Picture Archiving and Communication System), an HIS (Hospital Information System), a RIS (Radiology Information System), etc. In addition, the computer for machine learning can be connected via a LAN (Local Area Network) or other lines.

[0060] The above configuration is an example and is not limited thereto. For example, the sequence control circuit 29 may be incorporated into the medical data processing device 50. In addition, the sequence control circuit 29 and the processing circuit 51 may be mounted on the same substrate.

[0061] Hereinafter, an example of the magnetic resonance imaging apparatus 1 according to the present embodiment will be described.

[0062] As described above, the processing circuit 51 implements the output function 512 to perform the prescribed processing on the raw data and output the medical diagnosis data. Hereinafter, the prescribed processing for converting the raw data into the medical diagnosis data is referred to as the data generation processing. The parameters of the data generation processing can be set and changed by the user via the input interface 54, etc. Therefore, the parameters of the data generation processing may deviate due to the installation facilities or modification of the device, etc. This deviation becomes a statistical deviation of the input data of the machine learning, which becomes a cause of the deterioration of the accuracy of the machine learning.

[0063] Therefore, the processing circuit 51 of this embodiment performs data generation processing on raw data and outputs medical diagnostic data by realizing the output function 512, and does not perform part or all of the data generation processing, but outputs standardized medical data based on raw data for machine learning. According to the output function 512, part or all of the data generation processing, which is a cause of statistical deviation, is not performed on raw data, but medical data standardized for machine learning is output. Thus, the statistical deviation of the input data of machine learning is reduced or eliminated, and the accuracy of machine learning is improved.

[0064] The following describes the details of the processing of the processing circuit 51. In the following embodiments, it is assumed that the medical diagnostic data and the standardized medical data are MR images, and the data generation process includes a spatial transformation process, in other words, image reconstruction. The original data is k-space data. The original data can be two-dimensional k-space data or three-dimensional k-space data.

[0065] Figure 2 is a diagram showing the flow of data output processing of the processing circuit 51. Figure 2 As shown, the processing circuit 51 acquires the raw data collected by MR imaging from the receiving circuit 25 by implementing the acquisition function 511 (step SA1). In addition, the raw data collected by the receiving circuit 25 can be temporarily stored in the memory 52, and the processing circuit 51 can read the raw data of the processing object from the memory 52 after receiving the request for data generation processing.

[0066] When performing step SA1, the processing circuit 51 implements data generation processing (step SA2) on the raw data obtained in step SA1 by realizing the output function 512. The data generation processing SA2 includes a series of processing for generating medical diagnostic data, i.e., MR images for medical diagnosis from the raw data. For example, the data generation processing SA2 includes a spatial transformation processing SA21 and a post-processing SA22. The spatial transformation processing SA21 is image reconstruction from raw data to medical diagnostic data. In addition, the raw data for the data generation processing SA2 can be the raw k-space data collected by MR imaging, or it can be data after any data processing such as data compression processing, resolution decomposition processing, data interpolation processing, resolution synthesis processing, etc. is performed on the raw k-space data.

[0067] Image reconstruction may be any process as long as it converts raw data into an MR image, and may be, for example, Fourier transform or inverse Fourier transform, or a reconstruction method using machine learning.

[0068] Post-processing is a post-processing filter (post-filter) applied to MR images. The type of post-processing varies depending on the image quality and purpose required for medical diagnostic data. For example, denoising, smoothing filters, edge enhancement, segmentation, three-dimensional image processing, etc. are set as post-processing. Denoising is an image processing used to remove or reduce noise from various causes depicted in MR images. Smoothing filters are image processing that removes high-frequency components. Three-dimensional image processing is performed when the original data is three-dimensional k-space data. As three-dimensional image processing, volume rendering, surface rendering, MPR (Multi-planar reconstruction), pixel value projection method, etc. can be cited.

[0069] The processing circuit 51 performs spatial transformation processing SA21 on the raw data to generate an MR image, performs post-processing SA22 on the generated MR image, and outputs the MR image OA1 after the post-processing SA22 for medical diagnosis as medical diagnosis data (step SA3). In addition, the processing circuit 51 does not perform all post-processing SA22 on the MR image generated by the spatial transformation processing SA21, and outputs the MR image OA2 without post-processing for machine learning as standardized medical data (step SA4). For example, the processing circuit 51 performs spatial transformation processing SA21 on a single raw data to generate an MR image, copies the generated MR image, performs post-processing SA22 on one MR image to output the MR image OA1 for medical diagnosis, and outputs the other MR image as the MR image OA2 for machine learning. As another method, the processing circuit 51 can, for example, copy the original data, perform spatial transformation processing SA21 on the original data of one party to generate an MR image, perform post-processing SA22 on the generated MR image, and output an MR image OA1 for medical diagnosis, perform spatial transformation processing SA21 on the original data of the other party to generate an MR image, and output the generated MR image as an MR image OA2 for machine learning.

[0070] The MR image OA2 is outputted in parallel with the MR image OA1 at approximately the same time. For example, the MR image OA1 is displayed on the display 53 for medical diagnosis purposes of the user, etc., by the display control function 513 of the processing circuit 51. For example, the MR image OA2 is sent to a computer for machine learning through the communication interface 55 for machine learning processing. The machine learning processing may be parameter learning of a machine learning model or various processing using a learned model.

[0071] like Figure 2 As shown, the processing circuit 51 generates and outputs the MR image OA1 by performing the spatial transformation processing SA21 and the post-processing SF22 on the raw data, and generates and outputs the MR image OA2 by performing the spatial transformation processing SA21 on the raw data without performing all the post-processing SA22. That is, the MR image OA2 is generated without the post-processing SA22 accompanied by the statistical deviation caused by the modification of the installation facilities or the device, so it can be said to be standardized data. By using the standardized medical data without statistical deviation, that is, the MR image OA2, for machine learning processing, the accuracy of the machine learning processing can be improved. The processing circuit 51 automatically outputs the MR image OA2 as standardized medical data in parallel with the MR image OA1 as medical diagnosis data, so that the workflow of medical diagnosis can be maintained while improving the accuracy of machine learning.

[0072] Figure 32 is a diagram showing the flow of other data output processing of the processing circuit 51. Figure 3 As shown, the processing circuit 51 acquires the raw data collected by executing the data collection sequence from the receiving circuit 25 by realizing the acquisition function 511 (step SB1). Step SB1 is the same as step SA1, and thus its description is omitted.

[0073] After performing step SB1, the processing circuit 51 implements the data generation process on the raw data acquired in step SB1 by realizing the output function 512 (step SB2). Figure 3 The data generation process SB2 includes a spatial transformation process SB21 and a post-processing SB22. Figure 2 The post-processing SB22 includes two or more post-processings. For example, the post-processing SB22 includes a magnetic field distortion correction SB221 and a DICOM transformation SB222. The magnetic field distortion correction SB221 is an image processing for removing noise caused by distortion such as a static magnetic field, and is performed for both medical diagnosis data and standardized medical data. The DICOM transformation SB222 is performed only for medical diagnosis data.

[0074] The processing circuit 51 performs the spatial transformation processing SB21 on the raw data to generate an MR image, performs all SB221 and SB222 of the post-processing SB22 on the generated MR image, and outputs the DICOM-compliant MR image OB1 as medical diagnosis data (step SB3). In addition, the processing circuit 51 performs only a part (magnetic field distortion correction) SB221 of the post-processing SB22 on the MR image generated by the spatial transformation processing SB21, does not perform the other part (DICOM transformation) SB222, and outputs the non-DICOM-compliant MR image OB2 as standardized medical data (step SB4).

[0075] As mentioned above, according to Figure 3 The data generation process SB2 is performed for the standardized medical data, and the necessary post-processing SB221 in the post-processing SB22 is performed until 1. The necessary post-processing SB221 is not limited to the magnetic field distortion correction. It is also a process that should be performed for the standardized medical data, and as long as it is a correction that is not easy to produce statistical deviations associated with differences in the installation facilities or device modifications, it can be implemented as the necessary post-processing SB221. For example, as the necessary post-processing SB221, the gain adjustment of the receiving coil 47 can also be performed.

[0076] like Figure 3As shown, the processing circuit 51 performs spatial transformation processing SB21 and post-processing SB22 on the raw data to generate and output the MR image OB1, and performs spatial transformation processing SB21 and necessary post-processing SB221 of the post-processing SB22 on the raw data, and does not perform unnecessary post-processing SB222 of the post-processing SB22, and generates and outputs the MR image OB2. That is, the MR image OB2 is subjected to post-processing SB221 without statistical deviation, and is not generated through post-processing SB222 with statistical deviation, so it can be said to be standardized data. By using the MR image OB2 as standardized medical data without statistical deviation for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the MR image OB2 as standardized medical data in parallel with the MR image OB1 as medical diagnosis data, so the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0077] Figure 4 2 is a diagram showing the flow of other data output processing of the processing circuit 51. Figure 4 As shown, the processing circuit 51 acquires the raw data collected by executing the data collection sequence from the receiving circuit 25 by implementing the acquisition function 511 (step SC1). Step SC1 is the same as step SA1, and therefore its description is omitted.

[0078] After step SC1 , the processing circuit 51 implements the output function 512 to perform data generation processing on the raw data acquired in step SC1 (step SC2 ). Figure 4 The data generation process SC2 includes a spatial transformation process SC21 and a post-processing process SC22. Figure 2 The post-processing SC22 is an image processing capable of adjusting parameters. The types of post-processing SC22 can be the same as Figure 2 The same type of post-processing as post-processing SA22 is performed. The parameters of post-processing SC22 are set to different values ​​between the parameters for medical diagnosis data (hereinafter referred to as the first parameters) and the parameters for standardizing medical data (hereinafter referred to as the second parameters). The first parameter can be set to any value by the user so that the MR image has a quality suitable for medical diagnosis. The second parameter is set to a value after the input data is standardized for machine learning.

[0079] The processing circuit 51 performs the spatial transformation processing SC1 on the raw data to generate an MR image, performs the post-processing SC22 set as the first parameter on the generated MR image, and outputs the MR image OC1 as medical diagnosis data (step SC3). In addition, the processing circuit 51 performs the post-processing SC22 set as the second parameter on the MR image generated by the spatial transformation processing SC21, and outputs the MR image OC2 as standardized medical data (step SC4).

[0080] like Figure 4 As shown, the processing circuit 51 generates and outputs an MR image OC1 by performing a spatial transformation process SC21 and a post-processing SC22 of the first parameter on the raw data, and generates and outputs an MR image OC2 by performing a spatial transformation process SC21 and a post-processing SC22 of the second parameter on the raw data. That is, the MR image OC2 is generated by performing the post-processing SC22 of the second parameter unified for machine learning, so it can be said to be standardized data. By using the MR image OC2 as standardized medical data without statistical deviation for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the MR image OC2 as standardized medical data in parallel with the MR image OC1 as medical diagnosis data, so that the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0081] Figure 5 2 is a diagram showing the flow of other data output processing of the processing circuit 51. Figure 5 As shown, the processing circuit 51 acquires the raw data collected by executing the data collection sequence from the receiving circuit 25 by implementing the acquisition function 511 (step SD1). Step SD1 is the same as step SA1, and therefore its description is omitted.

[0082] After step SD1 is performed, the processing circuit 51 implements the data generation process on the raw data acquired in step SD1 by realizing the output function 512 (step SD2). Figure 5 The data generation process SD2 includes pre-processing SD21 and spatial transformation processing SD22. Pre-processing SD21 is a process other than spatial transformation processing performed on the original data. The parameters of pre-processing SD21 can be set to arbitrary values ​​by the user via the input interface 54, and it is a process that may produce statistical deviations due to differences in installation facilities or device modifications. Examples of pre-processing SD21 include denoising and smoothing filters. Spatial transformation processing SD22 and Figure 2 The spatial transformation processing SA21 is the same.

[0083] In step SD2, the processing circuit 51 performs pre-processing SD21 on the raw data, performs spatial transformation processing SD22 on the pre-processed raw data, and outputs the MR image OD1 as medical diagnosis data (step SD3). Alternatively, the processing circuit 51 performs spatial transformation processing SD22 without performing pre-processing SD21 on the raw data, and outputs the MR image OD2 as standardized medical data (step SD4).

[0084] like Figure 5 As shown, the processing circuit 51 generates and outputs the MR image OD1 by performing preprocessing SD21 and spatial transformation processing SD22 on the raw data, and generates and outputs the MR image OD2 by performing spatial transformation processing SD22 on the raw data without performing preprocessing SD21. That is, the MR image OD2 is not generated by the preprocessing SD21 with statistical deviation, so it can be said to be standardized data. By using the MR image OD2 as standardized medical data without statistical deviation for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the MR image OD2 as standardized medical data in parallel with the MR image OD1 as medical diagnosis data, so that the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0085] In addition, when a plurality of preprocessing processes are prepared and there is a preprocessing process that cannot become a statistical deviation, the preprocessing process may be performed to standardize the medical data. Figure 5 In the data output processing, it can also be performed after the spatial transformation processing SD22, such as Figure 2 , Figure 3 as well as Figure 4 Post-processing is performed as shown.

[0086] As described above, various data generation processes are performed by the processing circuit 51 .

[0087] In addition, in the above description, the original data may not be k-space data but mixed data. Mixed data is data that has been subjected to Fourier transform or inverse Fourier transform along at least one axis of k-space data. Mixed data may be original mixed data or data subjected to any data processing such as data compression processing, resolution decomposition processing, data interpolation processing, resolution synthesis processing, etc. on the original mixed data.

[0088] Next, the output format of the medical diagnosis data and the standardized medical data based on the realization of the output function 512 will be described.

[0089] Figure 6 is a diagram schematically showing the output form of medical diagnosis data and standardized medical data. Figure 6As shown, the processing circuit 51 can store and output the medical diagnostic data and the standardized medical data in a single DICOM (Digital Imaging and Communications in Medicine) file. In this case, the processing circuit 51 performs DICOM conversion on both the medical diagnostic data and the standardized medical data in the data generation process. DICOM conversion is the process of converting the data format of the MR image from a non-DICOM standard data format to a DICOM standard data format. The processing circuit 51 stores the DICOM-converted medical diagnostic data and the standardized medical data in the DICOM file. The method of storing the medical diagnostic data and the standardized medical data in the DICOM file is not particularly limited. For example, the processing circuit 51 outputs the medical diagnostic data and the standardized medical data as different series of the same entity. That is, the medical diagnostic data is output as a medical diagnosis series and the standardized medical data is output as a machine learning series.

[0090] Information indicating that the data is not for machine learning can be added to the medical diagnosis data, and information indicating that the data is for machine learning can be added to the standardized medical data. For example, the information can be each data in the form of a flag. For example, as a flag indicating that the data is not for machine learning, "NO" or "0" can be added, and as a flag indicating that the data is for machine learning, "YES" or "1" can be added. In addition, it is not limited to the flag form, and the information can also be included in the DICOM supplementary information of the medical diagnosis data and the standardized medical data.

[0091] In addition, the information indicating that the data is for machine learning and the information indicating that the data is not for machine learning may be text information expressing the same meaning. The text information may be stored in a DICOM dedicated tag of each data or directly attached to the actual data (e.g., MR image) of each data.

[0092] The processing circuit 51 supplies the standardized medical data to the computer for machine learning or the memory for machine learning while generating it. For example, when supplying the standardized medical data to the computer for machine learning, the processing circuit 51 sends the standardized medical data via the communication interface 55. The standardized medical data received by the computer for machine learning is provided for machine learning. At this time, the medical diagnosis data may not be sent to the computer for machine learning. The medical diagnosis data may be stored in the memory 52 of the magnetic resonance imaging device 1, or may be sent to the PACS via the communication interface 55. In addition, when the computer for machine learning is managed by a facility different from the installation facility of the magnetic resonance imaging device 1, only people with permission to use can access the standardized medical data, so the processing circuit 51 can encrypt the standardized medical data. As a result, in the so-called cross-site situation, the security related to the management of personal information, i.e., standardized medical data, can be improved to prevent the leakage of personal information.

[0093] The memory for machine learning may be the memory 52 mounted on the magnetic resonance imaging apparatus 1, a removable recording medium, or another computer. The standardized medical data stored in the memory for machine learning is appropriately supplied to the computer for machine learning for machine learning.

[0094] In addition, one or both of the medical diagnosis data and the standardized medical data may not be converted to the DICOM standard. If statistical deviation is generated by DICOM conversion, it is better not to convert the standardized medical data to DICOM.

[0095] Next, the display of medical diagnosis data and standardized medical data will be described.

[0096] Figure 7 FIG. 1 is a diagram showing an example of a GUI screen I1 related to data output processing. Figures 2 to 5 The data output process is displayed on the display 53 before the start of the process. Figure 7 As shown, a GUI button IB1 indicating whether or not to output medical diagnosis data and a GUI button IB2 indicating whether or not to output standardized medical data are displayed in the GUI screen I1. For example, when the GUI button IB1 is pressed via the input interface 54, etc., the output of medical diagnosis data is instructed, and when the GUI button IB2 is pressed via the input interface 54, etc., the output of standardized medical data is instructed.

[0097] When the GUI button IB1 and the GUI button IB2 are pressed, the processing circuit 51 Figures 2 to 5The data generation processing SA2, SB2, SC2 and SD2 shown in the figure outputs medical diagnosis data and standardized medical data. The standardized medical data is automatically sent to the computer for machine learning through the communication interface 55 at about the same time as the output. The standardized medical data received by the computer for machine learning is used for machine learning processing. For example, when the computer for machine learning performs automatic diagnosis based on medical images, the automatic diagnosis result is sent to the magnetic resonance imaging device 1. As an automatic diagnosis, there is automatic detection of lesions using CNN (Convolutional Neural Networks) and the like. In this case, the computer for machine learning applies a machine learning model to the standardized medical data to detect the lesion position and annotates the lesion detection position. Then, the computer for machine learning sends the annotated standardized medical data to the magnetic resonance imaging device 1. The annotated standardized medical data is received by the communication interface 55. The processing circuit 51 displays the medical diagnosis data and the annotated standardized medical data on the display 53 by realizing the display control function 513.

[0098] Figure 8 The medical diagnosis data I2 and the standardized medical data I3 annotated with the annotation M1 are displayed. Figure 7 FIG. 1 is an example of a GUI screen I1 of FIG. Figure 8 As shown, medical diagnosis data (MR image) I2 and standardized medical data (MR image) I3 are displayed side by side on the GUI screen I1. Standardized medical data I3 is the result data of machine learning sent from the computer used for machine learning. Annotations M1 are marked at the automatically detected positions of lesions in the standardized medical data I3. The user observes the standardized medical data I3 to evaluate the automatic detection results, and observes the medical diagnosis data I2 to make a diagnosis. By displaying the medical diagnosis data I2 and the standardized medical data I3, it is possible to refer to the standardized medical data I3 that reflects the high-precision machine learning results based on the standardized medical data with reduced statistical deviation, and to perform medical diagnosis based on the medical diagnosis data I2 that has been processed to generate data that the user prefers.

[0099] (Second Embodiment)

[0100] The medical data processing device of the second embodiment is a computer mounted on an X-ray computed tomography device. The second embodiment is described below. In addition, in the following description, the same reference numerals are marked on the components having substantially the same functions as those of the first embodiment, and the description is repeated only when necessary.

[0101] Fig. 9 2 is a diagram showing the structure of an X-ray computed tomography apparatus 1-2 according to the second embodiment. Fig. 9 Although a plurality of CT gantry 10 - 2 are depicted for convenience of explanation, the X-ray computed tomography apparatus 1 - 2 is typically equipped with only one gantry 10 - 2 .

[0102] like Fig. 9 As shown, the X-ray computed tomography apparatus 1-2 has a gantry 10-2, a diagnostic bed 30-2, and a medical data processing device (console) 50-2. The gantry 10-2 is a scanning device having a structure for performing X-ray CT photography on a subject P. The diagnostic bed 30-2 is a transport device for carrying the subject P as the object of X-ray CT photography and positioning the subject P. The medical data processing device 50-2 is a computer that controls the gantry 10-2. For example, the gantry 10-2 and the diagnostic bed 30-2 are arranged in an examination room, and the medical data processing device 50-2 is arranged in a control room adjacent to the examination room. The gantry 10-2, the diagnostic bed 30-2, and the medical data processing device 50-2 are connected by wire or wirelessly in a manner that allows them to communicate with each other.

[0103] like Fig. 9 As shown, the gantry 10 - 2 has an X-ray tube 11 , an X-ray detector 12 , a rotating frame 13 , an X-ray high voltage device 14 , a control device 15 , a wedge filter 16 , a collimator 17 , and a data acquisition circuit (DAS: Data Acquisition System) 18 .

[0104] The X-ray tube 11 generates X-rays. Specifically, the X-ray tube 11 includes a cathode that generates thermal electrons, an anode that receives thermal electrons flying from the cathode and generates X-rays, and a vacuum tube that holds the cathode and the anode. The X-ray tube 11 is connected to an X-ray high-voltage device 14 via a high-voltage cable. The filament current is supplied to the cathode by the X-ray high-voltage device 14. Thermal electrons are generated from the cathode by the supply of the filament current. A tube voltage is applied between the cathode and the anode by the X-ray high-voltage device 14. Due to the application of the tube voltage, thermal electrons fly from the cathode to the anode and collide with the anode, generating X-rays. The generated X-rays irradiate the subject P. Thermal electrons fly from the cathode to the anode, thereby flowing the tube current.

[0105] The X-ray detector 12 detects X-rays generated from the X-ray tube 11 and passing through the subject P, and outputs an electrical signal corresponding to the amount of X-rays detected to the DAS 18. The X-ray detector 12 has a structure in which a plurality of X-ray detection element rows are arranged in the slice direction (column direction) along the channel direction. The X-ray detector 12 is, for example, an indirect conversion type detector having a grid, a scintillator array, and a light sensor array. The scintillator array has a plurality of scintillators. The scintillator outputs light of an amount corresponding to the amount of incident X-rays. The grid is arranged on the X-ray incident surface side of the scintillator array and has an X-ray shielding plate that absorbs scattered X-rays. In addition, the grid is sometimes also referred to as a collimator (one-dimensional collimator or two-dimensional collimator). The light sensor array converts into an electrical signal corresponding to the amount of light from the scintillator. As a light sensor, for example, a photodiode is used.

[0106] The rotating frame 13 is a ring-shaped frame that supports the X-ray tube 11 and the X-ray detector 12 so as to be rotatable around the rotation axis Z. Specifically, the rotating frame 13 supports the X-ray tube 11 and the X-ray detector 12 opposite to each other. The rotating frame 13 is supported by a fixed frame (not shown) in a manner that allows rotation around the rotation axis Z. The rotating frame 13 is rotated around the rotation axis Z by the control device 15, thereby rotating the X-ray tube 11 and the X-ray detector 12 around the rotation axis Z. The image field of view (FOV: Field Of View) is set for the opening 29-2 of the rotating frame 13.

[0107] In addition, in this embodiment, the longitudinal direction of the rotation axis of the rotating frame 13 in the non-tilted state or the top plate 33-2 of the diagnostic bed 30-2 is defined as the Z direction, the direction perpendicular to the Z direction and horizontal to the ground is defined as the X direction, and the direction perpendicular to the Z direction and vertical to the ground is defined as the Y direction.

[0108] The X-ray high voltage device 14 has a high voltage generating device and an X-ray control device. The high voltage generating device has circuits such as a transformer and a rectifier, and generates a high voltage applied to the X-ray tube 11 and a filament current supplied to the X-ray tube 11. The X-ray control device controls the high voltage applied to the X-ray tube 11 and the filament current supplied to the X-ray tube 11. The high voltage generating device can be a transformer type or an inverter type. The X-ray high voltage device 14 can be set in the rotating frame 13 in the gantry 10-2, and can also be set in a fixed frame (not shown) in the gantry 10-2.

[0109] The wedge filter 26-2 adjusts the dose of X-rays irradiated to the subject P. Specifically, the wedge 26-2 attenuates the X-rays so that the dose of X-rays irradiated from the X-ray tube 11 to the subject P becomes a predetermined distribution. For example, a metal filter formed by processing a metal such as aluminum is used as the wedge 26-2. The wedge filter 26-2 is processed to have a predetermined target angle or a predetermined thickness. In addition, the wedge filter 26-2 is also called a bow-tie filter.

[0110] The collimator 27-2 limits the irradiation range of the X-rays after passing through the wedge filter 26-2. The collimator 27-2 slidably supports a plurality of lead plates that shield the X-rays, and adjusts the shape of the slits formed by the plurality of lead plates. The collimator 27-2 is also called an X-ray aperture.

[0111] DAS18 reads an electrical signal corresponding to the X-ray dose detected by the X-ray detector 12 from the X-ray detector 12, amplifies the read electrical signal, and integrates the electrical signal throughout the observation period, thereby collecting projection data having a digital value corresponding to the X-ray dose throughout the observation period. Projection data is a type of raw data. DAS18 is implemented, for example, by an ASIC equipped with a circuit element capable of generating projection data. The projection data generated by DAS18 is transmitted from a transmitter having a light emitting diode (LED) disposed on the rotating frame 13 to a receiver having a light emitting diode (LED) disposed on a non-rotating portion (e.g., a fixed frame) of the gantry 10-2 through optical communication, and is transmitted from the receiver to the medical data processing device 3-2. In addition, the method of transmitting the projection data from the rotating frame 13 to the non-rotating portion of the gantry 10-2 is not limited to the above-mentioned optical communication, and can be any method as long as it is a non-contact data transmission.

[0112] The diagnostic bed 30-2 includes a base 35-2 and a top plate 33-2. The base 35-2 is set on the ground. The base 35-2 is a structure that supports a support frame so that it can move in a vertical direction (Y direction) relative to the ground. The support frame is a frame set on the upper part of the base 35-2. The support frame supports the top plate 33-2 so that it can slide along the central axis Z. The top plate 33-2 is a flexible plate-shaped structure on which the subject P is placed. The diagnostic bed drive device is accommodated in the diagnostic bed 30-2. The diagnostic bed drive device is a motor or an actuator that generates power for moving the top plate 33-2 on which the subject P is placed. The diagnostic bed drive device operates according to the control of the control device 15 or the medical data processing device 50-2.

[0113] The control device 15 controls the X-ray high voltage device 14, DAS 18, and the diagnostic bed 30-2 in order to perform X-ray CT imaging according to the imaging control based on the processing circuit 51. The control device 15 has: a processing circuit including a CPU and a driving device including a motor and an actuator. The processing circuit has a processor including a CPU and a memory including a ROM and a RAM as hardware resources. The control device 15 controls the gantry 10-2 and the diagnostic bed 30-2, for example, according to an operation signal from an input interface 54 provided in the medical data processing device 50-2, the gantry 10-2, and the diagnostic bed 30-2. For example, the control device 15 controls the rotation of the rotating frame 13, the tilt of the gantry 10-2, and the movement of the top plate 33 and the diagnostic bed 30-2.

[0114] The medical data processing device (console) 50-2 is a computer having a processing circuit 51, a memory 52, a display 53, an input interface 54, and a communication interface 55. The medical data processing device 50-2 is the same as the medical data processing device 50 of the first embodiment, and thus description thereof is omitted.

[0115] Fig.10 2 is a diagram showing the flow of data generation processing by the processing circuit 51 of the second embodiment. Fig.10 As shown, the processing circuit 51 acquires the raw data (projection data) collected by X-ray CT imaging from the DAS 18 by implementing the acquisition function 511 (step SE1). Alternatively, the raw data collected by the DAS 18 may be temporarily stored in the memory 52, and upon receiving a request for data generation processing, the processing circuit 51 reads the raw data to be processed from the memory 52. ​​The raw data may be the raw projection data collected by X-ray CT imaging, or may be data obtained by subjecting the raw projection data to arbitrary data processing such as data compression processing, resolution decomposition processing, data interpolation processing, and resolution synthesis processing.

[0116] After performing step SE1, the processing circuit 51 implements the data generation process (step SE2) on the raw data obtained in step SE1 by realizing the output function 512. The data generation process SE2 includes pre-processing SE21 and successive approximate reconstruction SE22. Pre-processing SE21 is logarithmic transformation or offset correction. Successive approximate reconstruction SE22 is a kind of image reconstruction method for reconstructing CT images based on projection data. As the successive approximate reconstruction SE22, there are EM (expectation maximization) method, ART (algebraic reconstruction technique) method, etc. or their applications. In addition, the successive approximate reconstruction SE22 can also be a method in which an analytical reconstruction method such as FBP (Filtered BackProjection) is combined with the above-mentioned successive approximate reconstruction. As the successive approximate reconstruction SE22, a noise reduction method incorporating a statistical model, a scanner model, an anatomical model and / or machine learning can also be used. In the successive approximate reconstruction SE22, as an adjustable parameter, a parameter called a mixing ratio, which is a ratio of synthesis of an initial image and an update image, is known. In this embodiment, the mixing ratio for medical diagnosis data (hereinafter referred to as the first mixing ratio) and the mixing ratio for standardized medical data (hereinafter referred to as the second mixing ratio) are set to different values. The first mixing ratio can be set to an arbitrary value by the user. The second mixing ratio is set to a value standardized for input data of machine learning.

[0117] The processing circuit 51 performs pre-processing SE21 on the raw data, performs successive approximate reconstruction SE22 set to the first mixing ratio on the raw data after the pre-processing SE21, and outputs the CT image OE1 as medical diagnosis data (step SE3). In addition, the processing circuit 51 performs successive approximate reconstruction SE22 set to the second mixing ratio on the CT image after the pre-processing SE21, and outputs the CT image OE2 as standardized medical data (step SE4).

[0118] like Fig.10As shown, the processing circuit 51 generates and outputs a CT image OE1 by performing pre-processing SE21 and successive approximate reconstruction SE22 of a first mixing ratio on the raw data, and generates and outputs a CT image OE2 by performing pre-processing SE21 and successive approximate reconstruction SE22 of a second mixing ratio on the raw data. That is, the CT image OE2 is generated by performing successive approximate reconstruction SE22 of a second mixing ratio that is unified for machine learning, so it can be said to be standardized data. By using the CT image OE2 as standardized medical data without statistical deviation for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the CT image OE2 as standardized medical data in parallel with the CT image OE1 as medical diagnosis data, so that the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0119] In addition, parameters for medical diagnosis data and parameters for standardizing medical data may be prepared for other parameters other than the mixing ratio of the successive approximate reconstruction SE22. In addition, the X-ray computed tomography apparatus 1-2 can also be used in the same manner as the magnetic resonance imaging apparatus 1 of the first embodiment by replacing k-space data with projection data and replacing MR images with CT images. Figure 2-Figure 5 The data generation process shown.

[0120] (Third Embodiment)

[0121] The medical data processing device of the third embodiment is a computer mounted on an ultrasonic diagnostic device. The third embodiment is described below. In addition, in the following description, the same reference numerals are marked on the components having substantially the same functions as those of the first embodiment, and the description is repeated only when necessary.

[0122] Fig.11 FIG. 1 is a diagram showing the structure of an ultrasonic diagnostic apparatus 1-3 according to a third embodiment. Fig.11 As shown, the ultrasonic diagnostic apparatus 1-3 includes an ultrasonic probe 10-3 and a medical data processing apparatus (apparatus main body) 50-3.

[0123] The ultrasonic probe 10-3 performs ultrasonic scanning on a scanning area in a living body such as a patient, for example, according to the control of the medical data processing device 50-3. The ultrasonic probe 10-3 has, for example, a plurality of piezoelectric transducers, a matching layer, and a backing material. In the present embodiment, the ultrasonic probe 10-3 has, for example, a plurality of piezoelectric transducers arranged in a predetermined direction. The ultrasonic probe 10-3 is detachably connected to the medical data processing device 50-3.

[0124] The plurality of piezoelectric vibrators generate ultrasonic waves according to the driving signal supplied from the ultrasonic transmission circuit 57 of the medical data processing device 50-3. Thus, ultrasonic waves are transmitted from the ultrasonic probe 10-3 to the living body. When ultrasonic waves are transmitted from the ultrasonic probe 10-3 to the living body, the transmitted ultrasonic waves are sequentially reflected on the discontinuity surface of the acoustic impedance in the internal tissue of the living body and are received by the plurality of piezoelectric vibrators as reflected wave signals. The amplitude of the received reflected wave signal depends on the difference in acoustic impedance in the discontinuity surface that reflects the ultrasonic waves. In addition, when the transmitted ultrasonic pulse is reflected on the surface of a moving blood flow or a radiation absorbing tissue membrane, the reflected wave signal is subjected to frequency shift depending on the speed component of the moving body in the ultrasonic transmission direction according to the Doppler effect. The ultrasonic probe 10-3 receives the reflected wave signal from the living body and converts it into an electrical signal. The electrical signal is supplied to the medical data processing device 50-3.

[0125] Fig.11 The medical data processing device 50-3 shown is a computer that generates and displays an ultrasonic image based on the reflected wave signal received by the ultrasonic probe 10-3. Fig.11 As shown, the medical data processing device 50 - 3 includes an ultrasonic transmission circuit 57 , an ultrasonic reception circuit 56 , a processing circuit 51 , a memory 52 , a display 53 , an input interface 54 , and a communication interface 55 .

[0126] The ultrasonic transmission circuit 57 is a processor that supplies a driving signal to the ultrasonic probe 10-3. The ultrasonic transmission circuit 57 is implemented by, for example, a trigger generation circuit, a delay circuit, and a pulse circuit. The trigger generation circuit repeatedly generates a rate pulse for forming a transmitted ultrasonic wave at a predetermined rate frequency. The delay circuit assigns a delay time to each rate pulse generated by the trigger generation circuit, and the delay time is used to converge the ultrasonic wave generated from the ultrasonic probe 10-3 into a beam and determine the delay time of each piezoelectric vibrator required for the transmission directivity. The pulse generator circuit applies a driving signal (driving pulse) to a plurality of ultrasonic vibrators provided in the ultrasonic probe 10-3 at a timing based on the rate pulse. The delay time assigned to each rate pulse is arbitrarily changed by the delay circuit, thereby arbitrarily adjusting the transmission direction from the piezoelectric vibrator surface.

[0127] The ultrasonic receiving circuit 56 is a processor that performs various processing on the reflected wave signal received by the ultrasonic probe 10-3 and generates a received signal. The ultrasonic receiving circuit 56 is implemented, for example, by an amplifier circuit, an A / D converter, a receiving delay circuit, and an adder. The amplifier circuit amplifies the reflected wave signal received by the ultrasonic probe 10-3 for each channel and performs gain correction processing. The A / D converter converts the reflected wave signal after gain correction into a digital signal. The receiving delay circuit gives the digital signal a delay time required to determine the receiving directivity. The adder performs addition operation on multiple digital signals given delay time. Through the addition operation processing of the adder, a receiving signal that emphasizes the reflection component from the direction corresponding to the receiving directivity is generated. The receiving signal is a kind of raw data.

[0128] The processing circuit 51 is, for example, a processor that functions as the core of the ultrasonic diagnostic apparatus 1 to 3. The processing circuit 51 implements a function corresponding to a program by executing the program stored in the memory 52. ​​The processing circuit 51 has, for example, an acquisition function 511, an output function 512, a display control function 513, a B-mode processing function 514, and a Doppler mode processing function 515.

[0129] In the B-mode processing function 514, the processing circuit 51 generates B-mode data based on the reception signal received from the ultrasonic receiving circuit 56. Specifically, the processing circuit 51 performs envelope detection processing and logarithmic amplification processing on the reception signal received from the ultrasonic receiving circuit 56, and generates data (B-mode data) that expresses the signal strength in terms of brightness. The generated B-mode data is stored in a RAW data memory (not shown) as data on a two-dimensional ultrasonic scanning line (raster). B-mode data is a type of raw data.

[0130] In the Doppler mode processing function 515, the processing circuit 51 generates data (Doppler data) by frequency analysis of the received signal received from the ultrasonic receiving circuit 56, and the data (Doppler data) extracts the motion information of the blood flow in the ROI (Region Of Interest) set in the scanning area based on the Doppler effect. Specifically, the processing circuit 51 generates Doppler data as the motion information of the blood flow, which estimates the average velocity, average variance value, average energy value, etc. at a plurality of sampling points. The generated Doppler data is stored in a RAW data memory (not shown) as data on a two-dimensional ultrasonic scanning line.

[0131] The processing circuit 51 performs a color Doppler method of color flow mapping (CFM) in the Doppler mode processing function 515. In the CFM method, ultrasonic waves are transmitted and received multiple times on multiple scanning lines. The processing circuit 51 applies an MTI (Moving Target Indicator) filter to the data column at the same position to suppress the signal (clutter signal) from the stationary tissue or the slowly moving tissue, and extracts the signal from the blood flow. Then, the processing circuit 51 estimates the speed, variance or energy of the blood flow based on the extracted signal. Doppler data is a type of raw data.

[0132] Description of the memory 523 , the display 53 , the input interface 54 , the communication interface 55 , the acquisition function 511 , the output function 512 , and the display control function 513 will be omitted.

[0133] Fig.12 1 is a diagram showing the flow of data output processing by the processing circuit 51 of the third embodiment. Fig.12 As shown, the processing circuit 51 obtains the raw data collected by ultrasonic scanning from the ultrasonic receiving circuit 56 by implementing the acquisition function 511 (step SF1). In ultrasonic scanning, raw data of multiple frames of time series are collected in real time. The following processing steps are performed on all or each specified frame of the collected multiple frames. In addition, the raw data collected by the ultrasonic receiving circuit 56 may be temporarily stored in the memory 52, and after receiving the request for data generation processing, the processing circuit 51 reads the raw data of the processing object from the memory 52. ​​The raw data may be the original raw data collected by ultrasonic scanning, or it may be data that has been subjected to any data processing such as data compression processing, resolution decomposition processing, data interpolation processing, resolution synthesis processing, etc.

[0134] After performing step SF1, the processing circuit 51 implements data generation processing (step SF2) on the raw data obtained in step SF1 by implementing the output function 512. The data generation processing SF2 includes a series of processing for generating an ultrasonic image for medical diagnosis as medical diagnosis data from the raw data. For example, the data generation processing SF2 includes a spatial transformation processing SF21 and a post-processing SF22. The spatial transformation processing SF21 is a scan transformation processing for transforming B-mode data into a B-mode image representing the spatial distribution of B-mode data values, or transforming Doppler mode data into a Doppler mode image representing the spatial distribution of Doppler mode data values.

[0135] Post-processing is post-processing filtering (post-filter) performed on ultrasonic images. The types of post-processing vary depending on the image quality and purpose required for medical diagnostic data. For example, denoising, smoothing filters, edge enhancement, segmentation, and three-dimensional image processing are set as post-processing.

[0136] The processing circuit 51 performs the spatial transformation processing SF21 on the raw data to generate an ultrasonic image, performs the post-processing SF22 on the generated ultrasonic image, and outputs the ultrasonic image OF1 after the post-processing for medical diagnosis as medical diagnosis data (step SF3). In addition, the processing circuit 51 does not perform all the post-processing SF22 on the ultrasonic image generated by the spatial transformation processing SF21, and outputs the ultrasonic image OF2 without post-processing for machine learning as standardized medical data (step SF4).

[0137] like Fig.12 As shown, the processing circuit 51 generates and outputs an ultrasonic image OF1 by performing a spatial transformation process SF21 and a post-processing SF22 on the raw data, and generates and outputs an ultrasonic image OF2 by performing a spatial transformation process SF21 on the raw data without performing a post-processing SF22. That is, the ultrasonic image OF2 can be said to be standardized data because it is not generated by the post-processing SF22 with statistical deviations caused by facility or device modification. By using the ultrasonic image OF2 as standardized medical data without statistical deviations for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the ultrasonic image OF2 as standardized medical data in parallel with the ultrasonic image OF1 as medical diagnosis data, so that the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0138] In addition, the ultrasonic diagnostic apparatus 1-3 can perform the same operation as the magnetic resonance imaging apparatus 1 of the first embodiment by replacing k-space data with B-mode data or Doppler data and replacing MR images with ultrasonic images. Figure 2-Figure 5 The data generation process shown.

[0139] In addition, it is not necessary to generate and output standardized medical data for the raw data of all the collected frames. For example, the spatial transformation processing SF21 and the post-processing SF22 are performed on the raw data of all the collected frames, and the ultrasonic images OF1 of all the frames are displayed as animations on the display 53. When the user finds an ultrasonic image of a frame suitable for machine learning, he presses the freeze button. When the freeze button is pressed, the processing circuit 51 can determine the ultrasonic image OF1 with post-processing displayed on the display 53 at the time when the freeze button is pressed, and output the ultrasonic image OF2 before post-processing SF22 corresponding to the determined ultrasonic image OF1 with post-processing SF22 as standardized medical data.

[0140] (Other embodiments)

[0141] In the first, second and third embodiments described above, the data to be processed in the data generation process is the raw data collected by the medical image diagnostic device. However, the present embodiment is not limited thereto, and can also be applied to raw data collected by other medical image diagnostic devices, and can also be applied to waveform data collected by a biological information measuring device. In the case of waveform data as the processing object, no medical image is generated, and therefore, there is no need to perform spatial transformation processing as data generation processing. For waveform data, post-processing such as noise reduction processing, smoothing processing, and offset processing is performed as data generation processing. The parameters of these post-processing may produce fluctuations caused by the installation of facilities or device modifications, and therefore there are also parameters that produce statistical deviations.

[0142] The processing circuit 51 performs post-processing on the waveform data to generate and output the post-processed waveform data (hereinafter referred to as processed waveform data), and outputs the waveform data without performing all post-processing on the waveform data (hereinafter referred to as unprocessed waveform data). Figure 3 As shown, post-processing is classified into necessary post-processing performed on both medical diagnosis data and standardized medical data and arbitrary post-processing performed only on medical diagnosis data. Since the unprocessed waveform data is generated without post-processing with statistical deviation, it can be said to be standardized data. By using the unprocessed waveform data as standardized medical data without statistical deviation for machine learning processing, the accuracy of machine learning processing can be improved. The processing circuit 51 automatically outputs the unprocessed waveform data as standardized medical data in parallel with the processed waveform data as medical diagnosis data, so that the accuracy of machine learning can be improved while maintaining the workflow of medical diagnosis.

[0143] According to the above-mentioned embodiments, the medical data processing device 50 includes a processing circuit 51, which obtains medical data related to the subject, performs data generation processing on the medical data and outputs medical diagnosis data, and outputs standardized medical data based on the medical data that is standardized for machine learning without performing part or all of the data generation processing. According to the above-mentioned configuration, compared with the case where the medical diagnosis data is used as input data for machine learning, the generation conditions (collection conditions) of the input data for machine learning are stable, and the statistical deviation of the input data can be reduced or eliminated.

[0144] According to at least one of the embodiments described above, the accuracy of machine learning can be improved.

[0145] The term "processor" used in the above description refers to circuits such as CPU, GPU, or application-specific integrated circuit (ASIC), programmable logic device (for example, simple programmable logic device (SPLD), complex programmable logic device (CPLD), and field programmable gate array (FPGA)). The processor implements its functions by reading and executing the program stored in the storage circuit. In addition, instead of storing the program in the storage circuit, it is also possible to directly load the program into the circuit of the processor. In this case, the processor implements its functions by reading and executing the program loaded into the circuit. In addition, it is also possible not to execute the program, but to implement the function corresponding to the program by a combination of logic circuits. In addition, the processors of the present embodiment are not limited to the case where each processor is configured as a single circuit, and multiple independent circuits can be combined to form one processor to implement its functions. In addition, it is also possible to Figure 1 , Fig. 9 as well as Fig.11 Multiple components in a chip are integrated into one processor to realize its functions.

[0146] The embodiments of the present invention have been described, but the embodiments of the present invention are presented as examples and are not intended to limit the scope of the invention. The embodiments of the present invention can be implemented in various other ways, and various omissions, substitutions, and changes can be made without departing from the scope of the main purpose of the invention. The embodiments of the present invention and their modifications are included in the scope and main purpose of the invention, and are also included in the invention described in the claims and their equivalents.

Claims

1. A medical data processing device comprising: an acquisition unit that acquires medical data related to the subject; and an output unit that performs data generation processing on the medical data to output medical diagnosis data, and outputs standardized medical data based on the medical data that is standardized for use as learning data for machine learning without performing a part or all of the data generation processing on the medical data, The medical data is raw data related to the subject collected by a medical image diagnostic device. The output portion, performing a spatial transformation process on the raw data to generate transformed data, the spatial transformation process being the part of the data generation process, performing all post-processing on the transformed data to generate first post-processed data, and outputting the first post-processed data as the medical diagnosis data, wherein the post-processing is another part of the data generation process, A part of the post-processing is performed on the transformed data to generate second post-processed data, and the second post-processed data is output as the standardized medical data.

2. A medical data processing device comprising: an acquisition unit that acquires medical data related to the subject; and an output unit that performs data generation processing on the medical data to output medical diagnosis data, and outputs standardized medical data based on the medical data that is standardized for use as learning data for machine learning without performing a part or all of the data generation processing on the medical data, The medical data is raw data related to the subject collected by a medical image diagnostic device. The output portion, performing a spatial transformation process on the raw data to generate transformed data, the spatial transformation process being the part of the data generation process, performing post-processing of the transformed data according to the first parameter set by the user, and outputting the image-processed data as the medical diagnosis data, wherein the post-processing of the first parameter set by the user is another part of the data generation process, The post-processing of the second parameter standardized for machine learning is performed on the transformed data, and the standardized medical data is output.

3. The medical data processing device according to claim 1 or 2, wherein: The output unit encrypts the standardized medical data.

4. The medical data processing device according to claim 1 or 2, wherein: The output unit outputs a single DICOM file, ie, a Digital Imaging and Communications in Medicine file, storing the medical diagnosis data and the standardized medical data.

5. The medical data processing device according to claim 1 or 2, wherein: The output unit automatically sends the standardized medical data to a computer for machine learning.

6. The medical data processing apparatus according to claim 5, wherein: Also available: A display unit displays the medical diagnosis data and result data of the machine learning from the computer for machine learning.

7. The medical data processing device according to claim 1 or 2, wherein: Also available: The display unit displays a screen for selecting whether or not the medical diagnosis data can be output and / or whether or not the standardized medical data can be output.

8. The medical data processing device according to claim 1 or 2, wherein: The output unit adds information indicating that the medical diagnosis data is not to be used for machine learning and / or adds information indicating that the standardized medical data is to be used for machine learning.

9. A medical data processing method comprising the following steps: An acquisition step of acquiring medical data related to the subject; an output step of performing data generation processing on the medical data to output medical diagnosis data, and outputting standardized medical data based on the medical data that is standardized for use as learning data for machine learning without performing part or all of the data generation processing on the medical data, The medical data is raw data related to the subject collected by a medical image diagnostic device. In the output step, performing a spatial transformation process on the raw data to generate transformed data, the spatial transformation process being the part of the data generation process, performing all post-processing on the transformed data to generate first post-processed data, and outputting the first post-processed data as the medical diagnosis data, wherein the post-processing is another part of the data generation process, A part of the post-processing is performed on the transformed data to generate second post-processed data, and the second post-processed data is output as the standardized medical data.

10. A medical data processing method comprising the following steps: An acquisition step of acquiring medical data related to the subject; an output step of performing data generation processing on the medical data to output medical diagnosis data, and outputting standardized medical data based on the medical data that is standardized for use as learning data for machine learning without performing part or all of the data generation processing on the medical data, The medical data is raw data related to the subject collected by a medical image diagnostic device. In the output step, performing a spatial transformation process on the raw data to generate transformed data, the spatial transformation process being the part of the data generation process, performing post-processing of the transformed data according to the first parameter set by the user, and outputting the image-processed data as the medical diagnosis data, wherein the post-processing of the first parameter set by the user is another part of the data generation process, The post-processing of the second parameter standardized for machine learning is performed on the transformed data, and the standardized medical data is output.

Images