Ultrasonic time series data processing device and non-transitory storage medium

Abstract

The present application provides an ultrasonic time series data processing apparatus and a non-transitory storage medium. A Doppler processing section (18) or a beam data processing section (20) generates object time series data based on a received beam data string from a receiving section (16). An artifact prediction section (38) predicts a kind of artifact generated by the object time series data by inputting the object time series data to a learned artifact prediction learner (32). An artifact reduction section (40) performs artifact reduction processing based on the predicted kind of artifact. A display control section (24) notifies a user of a corrected ultrasonic image (62) on which the artifact reduction processing is performed or a prediction result of the artifact prediction section (38).

Description

[0001] Cross-referencing of related applications

[0002] This application claims priority to Japanese Patent Application No. 2022-080338, filed on May 16, 2022, the entire contents of which, including the description, claims, drawings and abstract, are incorporated herein by reference. Technical Field

[0003] This specification discloses an ultrasonic time series data processing device and an ultrasonic time series data processing program. Background Technology

[0004] In the past, ultrasound waves were repeatedly transmitted and received at the same location within the body being examined (in the same direction from the perspective of the ultrasound probe), and the resulting time-series received beam data strings, i.e., time-series data, were then visualized or analyzed.

[0005] As a product of visualizing time-series data, examples include: M-mode images in which time is represented by a horizontal axis representing time and depth by a vertical axis, with brightness lines extending along the time axis to represent tissue activity in the depth direction; or Doppler waveform images in which time is represented by a horizontal axis representing time and velocity by a vertical axis, based on the difference between the frequency of transmitted ultrasound and the frequency of received ultrasound to calculate the velocity of the examined site or the blood flow flowing within the examined site.

[0006] International Publication No. 2012 / 008173 discloses a method for analyzing time series data as follows: This method, while being non-invasive, accurately determines vascular diseases, particularly arteriosclerosis, vascular stenosis, and aneurysms. It involves sending ultrasound waves (vibration frequency f) to the pulsating vessel wall of a test subject and receiving reflected echoes whose vibration frequency changes to f0. Wavelet transform is applied to these reflected echoes to obtain a wavelet spectrum. Pattern decomposition is performed on the wavelet spectrum to obtain pattern-distinguished spectra. Inverse wavelet transform is then used to obtain pattern-distinguished waveforms along the time axis. Norm values ​​are calculated for each pattern, and by comparing these norm distributions with those obtained from normal individuals, the presence or prevalence of vascular diseases, or the incidence of specific vascular diseases, is determined.

[0007] However, in the ultrasonic image in which the time series data is visualized, a false image due to the time series data is sometimes generated. The so-called false image in this specification is an image that hinders the display of a desired waveform in the ultrasonic image in which the time series data is visualized. The false image is sometimes due to the operating conditions of the ultrasonic diagnostic apparatus, or due to the subject, or due to the electrical circuitry and the like within the ultrasonic diagnostic apparatus. For example, as the false image, there are: aliasing in a Doppler waveform image; clutter due to an unnecessary signal caused by body motion of the subject; a mirror effect generated due to saturation of Doppler data; or electrical noise generated from the electrical circuitry and the like of the ultrasonic diagnostic apparatus in the Doppler waveform image and the M-mode image. SUMMARY

[0008] The ultrasonic time series data processing apparatus disclosed in this specification aims to generate an ultrasonic image based on time series data in which a false image is reduced. Alternatively, the ultrasonic time series data processing apparatus disclosed in this specification aims to enable a user to easily grasp the kind of false image generated in the ultrasonic image based on time series data.

[0009] The ultrasonic time series data processing apparatus disclosed in this specification is characterized by comprising: a false image prediction section that inputs object time series data to a false image prediction learner, and predicts the kind of false image generated by the object time series data based on the output of the false image prediction learner with respect to the input, wherein the false image prediction learner has learned, using as learning data a combination of time series data representing a temporal change of a signal generated by repeatedly performing transmission and reception of an ultrasonic wave with respect to the same position within a subject site a plurality of times and information representing the kind of false image generated in an ultrasonic image based on the time series data, so as to predict and output the kind of false image generated by the input time series data based on the input time series data, the object time series data being the time series data generated by repeatedly performing transmission and reception of an ultrasonic wave with respect to the same position within an object subject site a plurality of times; a false image reduction section that performs false image reduction processing for reducing the false image based on the kind of false image predicted by the false image prediction section; and a display control section that causes a display section to display an ultrasonic image based on the object time series data to which the false image reduction processing is applied.

[0010] According to this structure, the false image reduction processing based on the predicted kind of false image is performed by the false image reduction section on the basis that the kind of false image generated by the object time series data in the ultrasonic image is predicted by the false image prediction section. Thereby, the false image is reduced in the ultrasonic image based on the object time series data.

[0011] Alternatively, as part of the artifact reduction process, the artifact reduction unit can generate an ultrasound image with reduced artifacts by inputting the object time series data into an image generation model that generates an ultrasound image with reduced artifacts based on the time series data that produces the artifacts.

[0012] Even when artifacts generated by object time-series data in ultrasound images cannot be reduced by changing the settings of the ultrasound diagnostic device, this structure allows for the generation of ultrasound images with reduced artifacts by inputting the object time-series data into the learned image generation model.

[0013] Alternatively, it may also include a notification unit that informs the user that the artifact reduction process has been performed.

[0014] Based on this structure, users can easily understand whether artifact reduction processing has been performed on the ultrasound image.

[0015] Alternatively, it may also include: a time series data generation unit that generates the object time series data; an artifact prediction unit that predicts in real time the types of artifacts generated by the object time series data generated by the time series data generation unit; an artifact reduction unit that performs artifact reduction processing in real time based on the artifact type prediction by the artifact prediction unit; and a display control unit that causes the display unit to display an ultrasound image based on the object time series data that has undergone artifact reduction processing in real time based on the artifact reduction processing.

[0016] In the ultrasonic time-series data processing apparatus disclosed in this specification, the artifact prediction unit can predict the types of artifacts generated by the object time-series data simply by inputting the object time-series data into the learned artifact prediction learner. That is, the computational load for predicting the types of artifacts generated by the object time-series data can be minimized, thereby enabling faster prediction of the types of artifacts generated by the object time-series data. Therefore, when displaying an ultrasonic image with reduced artifacts in real time as in this configuration, the delay in displaying the ultrasonic image relative to the acquisition time of the object time-series data can be reduced.

[0017] Furthermore, the ultrasound time-series data processing apparatus disclosed in this specification includes: an artifact prediction unit that inputs object time-series data into an artifact prediction learner and predicts the type of artifact generated by the object time-series data based on the output of the artifact prediction learner to the input, wherein the artifact prediction learner learns a combination of time-series data representing the time variation of a signal generated by repeatedly transmitting and receiving ultrasound waves at the same location within an examination site and based on reflected waves from the examination site or blood flowing within the examination site, and time-series data for generating artifacts in an ultrasound image based on the time-series data, and information representing the type of the artifact, as learning data, so that it can predict and output the type of artifact generated by the time-series data based on the input time-series data, wherein the object time-series data is the time-series data generated by repeatedly transmitting and receiving ultrasound waves at the same location within an examination site; and a notification unit that notifies a user of the prediction result of the artifact prediction unit.

[0018] Based on this structure, users can easily understand the types of artifacts produced by object time-series data in ultrasound images.

[0019] Alternatively, it may also include: a time series data generation unit that generates the object time series data; an artifact prediction unit that predicts the types of artifacts generated by the object time series data in real time based on the generation of the object time series data by the time series data generation unit; and a notification unit that notifies the user of the prediction results of the artifact prediction unit in real time based on the prediction of the types of artifacts by the artifact prediction unit.

[0020] In the ultrasound time-series data processing apparatus disclosed in this specification, even when a new ultrasound image with reduced artifacts is generated as an artifact reduction process, the ultrasound image with reduced artifacts can be generated simply by inputting the target time-series data into the image generation model. That is, the computational load for generating the corrected ultrasound image can be minimized, thereby enabling faster generation of the corrected ultrasound image. Furthermore, when changing the settings of the ultrasound diagnostic apparatus as an artifact reduction process, the change itself takes no time. Therefore, when the prediction results of the artifact types are notified to the user in real time, as in this configuration, the delay in notification of the prediction results relative to the time point when the target time-series data is acquired can be reduced.

[0021] Alternatively, the object being inspected and the inspected part are pulsating parts, and the object time series data and the learning time series data are time series data corresponding to the same period in the pulsation cycle of the object being inspected and the inspected part.

[0022] According to this structure, by making the periods of the learning time series data and the object time series data in the pulsation cycle the same, the output accuracy of the artifact prediction learner can be improved, thereby improving the prediction accuracy of the types of artifacts represented by the object time series data.

[0023] Furthermore, the ultrasound time-series data processing program disclosed in this specification enables a computer to function as an artifact prediction unit that inputs object time-series data into an artifact prediction learner and predicts the types of artifacts generated by the object time-series data based on the output of the artifact prediction learner for the input. The artifact prediction learner uses time-series data representing the temporal changes of a signal generated by repeatedly transmitting and receiving ultrasound waves at the same location within an examined area, based on reflected waves from the examined area or blood flowing within the examined area, and serves as a time-series data for generating artifacts in ultrasound images based on the time-series data. The learning process involves using a combination of time-series data and information representing the type of artifact as learning data to predict and output the type of artifact generated by the time-series data based on the input time-series data. The object time-series data is generated by repeatedly transmitting and receiving ultrasound waves at the same location within the inspected part of the object. An artifact reduction unit performs artifact reduction processing to reduce the artifact based on the type of artifact predicted by the artifact prediction unit. A display control unit causes a display unit to display an ultrasound image based on the object time-series data that has undergone the artifact reduction processing.

[0024] Furthermore, the ultrasound time-series data processing program disclosed in this specification enables a computer to function as follows: an artifact prediction unit that inputs object time-series data into an artifact prediction learner and predicts the type of artifact generated by the object time-series data based on the output of the artifact prediction learner to the input, wherein the artifact prediction learner learns a combination of time-series data representing the time variation of a signal generated by repeatedly transmitting and receiving ultrasound waves at the same location within an examined area and based on reflected waves from the examined area or blood flowing within the examined area, and time-series data used as learning data for generating artifacts in ultrasound images based on the time-series data, and information representing the type of the artifact, so that the type of artifact generated by the time-series data is predicted and output based on the input time-series data, the object time-series data being generated by repeatedly transmitting and receiving ultrasound waves at the same location within an examined area; and a notification unit that notifies the user of the prediction results of the artifact prediction unit.

[0025] The ultrasonic time-series data processing apparatus disclosed in this specification can generate ultrasonic images based on time-series data with reduced artifacts. Alternatively, the ultrasonic time-series data processing apparatus disclosed in this specification allows users to easily determine the types of artifacts generated in ultrasonic images based on time-series data. Attached Figure Description

[0026] Figure 1 This is a block diagram of the ultrasonic diagnostic device according to this embodiment.

[0027] Figure 2 This is a conceptual diagram illustrating the relationship between received beam data and received frame data.

[0028] Figure 3 This is a conceptual diagram representing the learning process of an artifact prediction learner.

[0029] Figure 4 This is a conceptual diagram representing the prediction processing using an artifact prediction learner.

[0030] Figure 5 This is an example of a notification screen that displays the prediction results from the artifact prediction unit.

[0031] Figure 6 This is a first example of an ultrasound image with reduced artifacts.

[0032] Figure 7 This is a second example of an ultrasound image with reduced artifacts.

[0033] Figure 8 This is a flowchart illustrating the processing flow of the ultrasonic diagnostic apparatus according to this embodiment. Detailed Implementation

[0034] Figure 1 This is a block diagram illustrating the ultrasound diagnostic apparatus 10, which is an ultrasound time-series data processing device according to this embodiment. The ultrasound diagnostic apparatus 10 is installed in medical institutions such as hospitals and is a medical device used during ultrasound examinations.

[0035] The ultrasound diagnostic device 10 can operate in multiple modes, including B-mode, Doppler mode, and M-mode. B-mode is a mode in which a tomographic image (B-mode image) is generated and displayed, converting the amplitude intensity of the reflected wave from the scanning surface into brightness, based on received frame data composed of multiple received beam data obtained from the scanning ultrasound beam (transmitted beam). Doppler mode is a mode in which a waveform (Doppler waveform) representing the movement velocity of tissue along an observation line set within the subject is generated and displayed, based on the frequency difference between the transmitted and reflected waves along that observation line. Doppler modes may include continuous wave mode, pulsed Doppler mode, color Doppler mode, or tissue Doppler mode. M-mode 2 is a mode in which an M-mode image characterizing tissue movement along an observation line set within the subject is generated and displayed, based on received beam data corresponding to that observation line. This embodiment specifically focuses on the case where the ultrasound diagnostic device 10 operates in Doppler mode or M-mode.

[0036] The ultrasonic probe, or probe 12, is a device for transmitting ultrasonic waves and receiving reflected waves. Specifically, the probe 12 is in contact with the surface of the subject, transmitting ultrasonic waves into the subject and receiving reflected waves reflected from the tissues within the subject. The probe 12 contains an array of vibrating elements, comprising multiple vibrating elements. A transmission signal, as an electrical signal, is supplied to each vibrating element in the array from the transmitting unit 14 (described later), thereby generating an ultrasonic beam (transmission beam). Furthermore, each vibrating element in the array receives reflected waves from the subject, converts the reflected waves into a receiving signal, as an electrical signal, and transmits it to the receiving unit 16 (described later).

[0037] When transmitting ultrasound, the transmitting unit 14 supplies multiple transmission signals in parallel to the probe 12 (specifically, the vibrating element array) in accordance with the control of the processor 36 described later. Thus, ultrasound is transmitted from the vibrating element array.

[0038] In Doppler mode or M mode, the transmitter 14 supplies a transmission signal to the probe 12, causing the probe 12 to repeatedly transmit a transmission beam to the same location within the examination area of ​​the subject, as determined by the user (such as a physician or examination technician). In other words, the transmitter 14 supplies a transmission signal to the probe 12, causing the probe 12 to repeatedly transmit a transmission beam in the direction toward the same location within the examination area. Additionally, in B mode, the transmitter 14 supplies a transmission signal to the probe 12, causing the transmission beam transmitted from the probe 12 to perform electronic scanning within the scanning plane. Alternatively, time-division scanning can be performed, allowing the transmission beam to be repeatedly transmitted toward the same location determined by the user during the electronic scanning of the transmission beam within the scanning plane.

[0039] When receiving reflected waves, the receiving unit 16 receives multiple received signals from the probe 12 (specifically, an array of vibrating elements) in parallel. In the receiving unit 16, a phase modulation addition operation (delay addition operation) is performed on the multiple received signals, thereby generating received beam data.

[0040] In Doppler mode or M mode, the probe 12 repeatedly transmits a beam at the same location within the examined area. The receiver 16 receives multiple reflected waves from the examined area or blood flowing within it, and generates a time-series received beam data string based on these reflected waves. In B mode, the receiver 16 constructs received frame data using multiple received beam data arranged in the scanning direction.

[0041] Figure 2 This is a conceptual diagram illustrating the relationship between received beam data (BD) and received frame data (F). In Doppler or M-mode, ultrasonic waves are transmitted and received towards a user-specified location (direction). This generates multiple time-series received beam data (DB) (i.e., received beam data strings). The received beam data DB contains information representing the intensity and frequency of reflected waves from various depths. In B-mode, the transmitted beam is scanned along a scanning direction θ, and received frame data (F) is generated by arranging multiple received beam data along the scanning direction θ.

[0042] In Doppler mode, the received beam data string is sent to the Doppler processing unit 18; in M ​​mode, the received beam data string is sent to the beam data processing unit 20.

[0043] Back Figure 1 In Doppler mode, the Doppler processing unit 18 generates Doppler data as time-series data, representing the time-varying velocity of blood flowing through the examined area or within the examined area, based on the received beam data string from the receiving unit 16. Specifically, the Doppler processing unit 18 multiplies each received beam data by a reference frequency and performs orthogonal detection to extract the Doppler frequency shift through a low-pass filter, sampling gate processing (in the case of pulse Doppler mode) to extract only the signal at the location of the sampling volume, signal A / D conversion, and frequency analysis based on the high-speed Fourier transform (FFT) method to generate Doppler data. The generated Doppler data is then sent to the image generation unit 22 and the processor 36. In Doppler mode, the receiving unit 16 and the Doppler processing unit 18 function as the time-series data generation unit.

[0044] In M-mode, the beam data processing unit 20 performs various signal processing operations on the received beam data string from the receiving unit 16, including gain correction, logarithmic amplification, and filtering. The processed received beam data string is then sent to the image generation unit 22 and the processor 36. In this embodiment, in M-mode, the processed received beam data string from the beam data processing unit 20 corresponds to time-series data representing the temporal change in the position of the detected region. In this case, the receiving unit 16 and the beam data processing unit 20 correspond to the time-series data generation unit. Furthermore, in B-mode, the beam data processing unit 20 also performs the aforementioned signal processing on the received frame data from the receiving unit 16.

[0045] The image generation unit 22 is composed of a digital scan converter and has coordinate transformation function, pixel interpolation function, frame rate transformation function, etc.

[0046] In Doppler mode, the image generation unit 22 generates a Doppler waveform image based on Doppler data from the Doppler processing unit 18. The Doppler waveform is a waveform shown on a two-dimensional plane of time and velocity, representing the temporal change in the velocity of the examined part or the blood flowing within the examined part on the observation line corresponding to the received beam data string.

[0047] In M-mode, the image generation unit 22 generates an M-mode image based on the received beam data string from the beam data processing unit 20. The M-mode image is a waveform shown on a two-dimensional plane of time and depth, representing the temporal change of the position of the inspected part on the observation line corresponding to the received beam data string.

[0048] In ultrasound images (Doppler waveform images or M-mode images) generated by the image generation unit 22 based on time-series data (Doppler data or received beam data strings after signal processing), artifacts sometimes occur. As described above, the so-called artifacts in this specification are images in ultrasound images that image the time-series data, which obstruct the display of the desired waveform. Examples include: folding and mirroring effects caused by the operating conditions (settings) of the ultrasound diagnostic device 10; noise caused by the subject; or electrical noise caused by electrical circuits within the ultrasound diagnostic device. For example, in the case of movement caused by heart valves, noise has a wide range of components from high speed to low speed, appearing as bright lines approximately parallel to the vertical axis in a Doppler waveform image with time on the horizontal axis and velocity on the vertical axis. Furthermore, in the case of movement caused by the heart wall, noise has low-speed components with high signal intensity, appearing as bright lines extending laterally near lines with velocity of 0 in a Doppler waveform image with time on the horizontal axis and velocity on the vertical axis. Of course, the types of artifacts are not limited to those mentioned above.

[0049] In addition, in B mode, the image generation unit 22 generates a B mode image based on the received frame data from the beam data processing unit 20, in which the amplitude (intensity) of the reflected wave is characterized by brightness.

[0050] The display control unit 24 displays various images generated by the image generation unit 22, such as Doppler waveform images, M-mode images, or B-mode images, on a display unit 26, which is composed of, for example, a liquid crystal panel. Furthermore, the display control unit 24 displays the prediction results of the artifact prediction unit 38 (described later) or the processing results of the artifact reduction unit 40 (described later), i.e., the ultrasonic image, on the display unit 26.

[0051] Furthermore, each of the transmitting unit 14, receiving unit 16, Doppler processing unit 18, beam data processing unit 20, image generation unit 22, and display control unit 24 is composed of one or more processors, chips, electrical circuits, etc. Each unit can be implemented through the cooperation of hardware and software.

[0052] The input interface 28 may be composed of, for example, buttons, a trackball, or a touch panel. The input interface 28 is used to input user instructions to the ultrasound diagnostic device 10.

[0053] The memory 30 is configured with components such as HDD (Hard Disk Drive), SSD (Solid State Drive), eMMC (embedded Multi Media Card), ROM (Read Only Memory), or RAM (Random Access Memory). The memory 30 stores ultrasound time-series data processing programs used to operate various parts of the ultrasound diagnostic device 10. Alternatively, the ultrasound time-series data processing programs can also be stored on computer-readable non-transitory storage media such as USB (Universal Serial Bus) memory or CD-ROM. The ultrasound diagnostic device 10 or other computers can read and execute the ultrasound time-series data processing programs from such storage media. Furthermore, as... Figure 1 As shown, the artifact prediction learner 32 and the image generation model 34 are stored in the memory 30.

[0054] The artifact prediction learner 32 is composed of learning models such as RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), CNN (Convolutional Neural Network), or DQN (Deep Q-Network) which utilizes deep reinforcement learning algorithms. The artifact prediction learner 32 learns from time-series data generated by repeatedly transmitting and receiving ultrasound waves at the same location within the examined area, based on reflected waves from the examined area or blood flowing within the examined area, and also learns from time-series data that generates artifacts in ultrasound images based on this time-series data, combined with information (labels) indicating the type of artifact, so that it can predict and output the type of artifact generated by the time-series data based on the input time-series data.

[0055] Figure 3 This is a conceptual diagram illustrating the learning process of the artifact prediction learner 32. For example, the artifact prediction learner 32 is input with time-series data representing the types of artifacts, such as "wrinkles," generated in a Doppler waveform image. In this case, the artifact prediction learner 32 predicts and outputs the artifacts generated by the time-series data. By setting activation functions such as the softmax function in the final stage (output layer) of the artifact prediction learner 32, it can output the probabilities of the time-series data for each type of artifact as output data. The computer performing the learning process calculates the error between the output data and the label (in this case, "wrinkles") on the time-series data using a given loss function, and adjusts the parameters of the artifact prediction learner 32 (e.g., the weights and biases of each neuron) to reduce this error. By repeating this learning process, the artifact prediction learner 32 can predict and output the types of artifacts generated by the input time-series data.

[0056] The site of examination that becomes the object of time-series data for learning is sometimes a pulsating site. In this case, the time-series data for learning can be data corresponding to a given period in the pulsation cycle of the site of examination. For example, the electrocardiogram waveform of the subject can be obtained from an electrocardiograph equipped with the subject, and the time-series data of the received beam data string obtained based on the period between R waves in the electrocardiogram waveform can be used as the time-series data for learning.

[0057] In this embodiment, the learning time series data is the data before image formation (equivalent to the Doppler data and received beam data string mentioned above). However, the learning time series data can also be a Doppler waveform image or an M-mode image that is imaged based on the Doppler data and received beam data string (specifically, data that quantifies the features of the Doppler waveform image or the M-mode image).

[0058] In addition, Figure 3 The training data may include time-series data that does not produce artifacts (normal time), but it is also possible to exclude normal time-series data from the training data. Furthermore, the artifact prediction learner 32 can prepare different learners for each type of artifact. In this case, each learner learns such that the output of the input time-series data shows the probability of the corresponding type of artifact.

[0059] In this embodiment, the artifact prediction learner 32 is learned by a computer other than the ultrasound diagnostic device 10, and the learned artifact prediction learner 32 is stored in the memory 30. However, time-series data obtained from the ultrasound diagnostic device 10 can also be used as learning time-series data, and the learning processing of the artifact prediction learner 32 can be performed in the ultrasound diagnostic device 10. In this case, the processor 36 functions as a learning processing unit that performs the learning processing of the artifact prediction learner 32.

[0060] Image generation model 34 is a learning model that takes latent variables as input and generates two-dimensional images based on those latent variables. Image generation model 34 is, for example, composed of GANs (Generative Adversarial Networks). A GAN consists of an image generator that generates two-dimensional images based on latent variables and an image recognizer that identifies whether an image of an object is a generated image produced by the image generator. The image recognizer learns to identify whether an image of an object is a generated image with higher accuracy. For example, the image recognizer uses a combination of generated images and information indicating that they are generated images (labels), and a combination of real images (images that are not generated images) and information indicating that they are real images, as learning data. On the other hand, the image generator learns to generate images that closely resemble real objects, thus deceiving the image recognizer (the image recognizer misidentifies). For example, the image generator learns to make generated images based on latent variables judged as real images by the image recognizer. A well-learned GAN (specifically, the image generator contained within the GAN) can generate images that are even closer to real objects.

[0061] In this embodiment, the image generation model 34 learns to use time-series data that generates artifacts as latent variables (input data) and generates an ultrasound image (specifically a Doppler waveform image or an M-mode image) based on the time-series data with reduced artifacts. For example, the image generation model 34 learns to use time-series data that generates clutter as input data to generate an ultrasound image with reduced clutter.

[0062] Back Figure 1 The processor 36 comprises at least one of a general-purpose processing device (such as a CPU (Central Processing Unit)) and a special-purpose processing device (such as a GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or programmable logic device). The processor 36 may also be not based on a single processing device, but rather constituted through the cooperation of multiple processing devices located in physically separate locations. Figure 1 As shown, the processor 36 performs the functions of the artifact prediction unit 38 and the artifact reduction unit 40 according to the ultrasonic time series data processing program stored in the memory 30.

[0063] Figure 4 This is a conceptual diagram illustrating the prediction processing using the artifact prediction learner 32. The artifact prediction unit 38 inputs time-series data (referred to as "object time-series data" in this specification) obtained by repeatedly transmitting and receiving ultrasound waves at the same location within the object being inspected to the learned artifact prediction learner 32. The object being inspected is the part of the object being inspected that receives and receives ultrasound waves to obtain object time-series data (time-series data of the predicted object that determines the type of artifacts generated). As described above, the object time-series data is generated by the receiving unit 16 and the Doppler processing unit 18 (in the case of Doppler mode), or by the receiving unit 16 and the beam data processing unit 20 (in the case of M-mode). Furthermore, the object time-series data can be a Doppler waveform image or an M-mode image (specifically, data quantified from the features of the Doppler waveform image and the M-mode image) that is visualized based on Doppler data and received beam data streams.

[0064] The artifact prediction learner 32 predicts the types of artifacts generated by the input object time series data based on the input object time series data, and outputs output data representing the prediction results. The artifact prediction unit 38 predicts the types of artifacts generated by the object time series data based on the output data of the artifact prediction learner 32. When the artifact prediction learner 32 outputs the probabilities generated by the object time series data for multiple types of artifacts as output data, the artifact prediction unit 38 can predict the probabilities generated by the object time series data for multiple types of artifacts respectively.

[0065] When multiple artifact prediction learners 32 are prepared according to the types of artifacts, the artifact prediction unit 38 sequentially inputs the object time series data into the multiple artifact prediction learners 32, and predicts the probability of the object time series data based on the output data of each of the multiple artifact prediction learners 32 for multiple types of artifacts.

[0066] When the examined part of the target and the examined part of the target time series data used for learning are pulsating parts, the target time series data and the learning time series data can be time series data corresponding to the same period in the pulsation cycle of the examined part and the examined part. For example, if the learning time series data is time series data based on the received beam data string obtained during the period between R waves in an electrocardiogram waveform, the artifact prediction unit 38 can also set the target time series data to be time series data based on the received beam data string obtained during the period between R waves in an electrocardiogram waveform. By making the periods of the learning time series data and the target time series data in the pulsation cycle the same, the output accuracy of the artifact prediction learner 32 can be improved. That is, the prediction accuracy of the types of artifacts generated by the target time series data of the artifact prediction unit 38 can be improved.

[0067] The artifact reduction unit 40 performs artifact reduction processing to reduce the type of artifact predicted by the artifact prediction unit 38.

[0068] When artifacts generated by the time series data of an object in an ultrasound image can be reduced by changing the operating conditions (settings) of the ultrasound diagnostic device 10, the artifact reduction unit 40 performs a process of changing the settings of the ultrasound diagnostic device 10 as an artifact reduction process.

[0069] For example, if the artifact type predicted by the artifact prediction unit 38, i.e., the artifact generated by the object time series data, is "wrinkling," the artifact reduction unit 40 performs zero-point drifting, shifting the zero-hertz line of the Doppler waveform up and down, to eliminate the wrinkling. Alternatively, the artifact reduction unit 40 adjusts the PRF (Pulse Repetition Frequency) to eliminate wrinkling. Furthermore, if the artifact type predicted by the artifact prediction unit 38 is "mirror effect," the artifact reduction unit 40 adjusts the gain to reduce the mirror effect. Moreover, if the artifact type predicted by the artifact prediction unit 38 is "electrical noise," the artifact reduction unit 40 adjusts the gain to reduce electrical noise.

[0070] On the other hand, if artifacts generated in the ultrasound image by the object time-series data cannot be reduced by changing the settings of the ultrasound diagnostic device 10, the artifact reduction unit 40 inputs the object time-series data into the learned image generation model 34 for artifact reduction processing. The image generation model 34 outputs an ultrasound image with reduced artifacts based on the object time-series data, which is input as a latent variable. This is how an ultrasound image with reduced artifacts is generated. Furthermore, the input to the image generation model 34 can be a Doppler waveform image or an M-mode image visualized based on the object time-series data (specifically, data quantified from the features of the Doppler waveform image or the M-mode image).

[0071] The display control unit 24 notifies the user of the prediction results of the artifact prediction unit 38. That is, the display control unit 24 functions as a notification unit. In addition, in this embodiment, the prediction results of the artifact prediction unit 38 are displayed on the display 26 by the display control unit 24 as described below, but it is also possible to notify the user of the prediction results of the artifact prediction unit 38 by means of sound output or the like.

[0072] Figure 5 This is an example diagram of a notification screen 50 displayed on the display 26, notifying the artifact prediction unit 38 of its prediction results. The notification screen 50 displays an ultrasound image 52 generated based on object time-series data, and the prediction result 54 of the artifact prediction unit 38, which predicts the types of artifacts generated by the object time-series data. Furthermore, the notification screen 50 is a Doppler mode screen, displaying a Doppler waveform image as the ultrasound image 52; of course, in M-mode, it displays an M-mode image as the ultrasound image 52.

[0073] When the artifact prediction unit 38 predicts the probability of the object time series data conforming to multiple types of artifacts, the display control unit 24 can notify the user of the probability of the object time series data conforming to multiple types of artifacts.

[0074] In this way, by notifying the user of the prediction results from the artifact prediction unit 38, the user can easily understand whether artifacts have occurred in the ultrasound image 52 based on the object's time-series data, and what types of artifacts have occurred. This is especially beneficial for users who are not accustomed to viewing ultrasound images 52, as they may find it difficult to determine whether artifacts have occurred in the displayed ultrasound image 52. For such users, a prediction of whether artifacts have occurred in the displayed ultrasound image 52, or the type of artifact, is particularly helpful.

[0075] Figure 6 as well as Figure 7 This is a diagram showing a display screen 60 displaying a corrected ultrasound image 62 with reduced artifacts, displayed on the display 26. When the artifact reduction unit 40 performs the aforementioned artifact reduction processing, the display control unit 24 causes the display 26 to display the corrected ultrasound image 62 based on object time-series data, which has undergone artifact reduction processing.

[0076] For example, in Figure 6 In the example, the display control unit 24 displays a corrected ultrasonic image 62, which is an ultrasonic image based on time-series data of the object that produced the wrinkles, and which has had the wrinkles eliminated by zero-point drift. In the case where image generation by the image generation model 34 is performed as an artifact reduction process, such as... Figure 7 As shown, the display control unit 24 displays the corrected ultrasound image 62 generated by the image generation model 34.

[0077] In addition to correcting the ultrasound image 62, the display control unit 24 can also notify the user that artifact reduction processing has been performed on the corrected ultrasound image 62. Furthermore, the display control unit 24 can notify the user of the types of artifacts that have been reduced through artifact reduction processing. For example, in Figure 6 In the corrected ultrasound image 62 shown, since the wrinkles have been eliminated, the display control unit 24 displays a notification message 64 indicating that wrinkle correction (specifically, zero-point drift) has been performed on the corrected ultrasound image 62. Similarly, in Figure 7 In the corrected ultrasound image 62 shown, since the noise has been reduced, the display control unit 24 displays a notification message 64 indicating that the noise has been reduced.

[0078] Therefore, users can easily understand that the displayed corrected ultrasound image 62 is an image that has undergone artifact reduction processing.

[0079] As described above, in this embodiment, the artifact prediction unit 38 can predict the types of artifacts generated by the object time series data by simply inputting the object time series data into the learned artifact prediction learner 32. That is, the computational load for predicting the types of artifacts generated by the object time series data can be kept low, thereby enabling faster prediction of the types of artifacts generated by the object time series data.

[0080] Therefore, the artifact prediction unit 38 can predict the types of artifacts generated by the object time series data in real time based on the generation of the object time series data, and the display control unit 24 can notify the user of the prediction results of the artifact prediction unit 38 in real time based on the prediction of the types of artifacts generated by the object time series data by the artifact prediction unit 38. According to this embodiment, even with such real-time processing, the prediction results of the artifact prediction unit 38 can be smoothly notified to the user without causing delays or the like.

[0081] Furthermore, even when the artifact reduction unit 40 generates a new corrected ultrasound image as an artifact reduction process, the artifact reduction unit 40 can generate a corrected ultrasound image with reduced artifacts simply by inputting the object time series data into the image generation model 34. That is, the computational load for generating the corrected ultrasound image can be minimized, thereby enabling faster generation of the corrected ultrasound image. Additionally, when changing the settings of the ultrasound diagnostic device 10 as an artifact reduction process, the change itself takes very little time.

[0082] Therefore, the artifact prediction unit 38 can predict the types of artifacts generated by the object time series data in real time based on the generation of the object time series data, the artifact reduction unit 40 can perform artifact reduction processing in real time based on the artifact type prediction of the artifact prediction unit 38, and the display control unit 24 can make the display 26 display the corrected ultrasound image 62 based on the object time series data in real time based on the artifact reduction processing. According to this embodiment, even with such real-time processing, the display 26 can smoothly display the corrected ultrasound image 62 without causing delays or the like.

[0083] The following is in accordance with Figure 8 The flowchart shown illustrates the processing flow of the ultrasonic diagnostic apparatus 10 according to this embodiment.

[0084] In step S10, the ultrasound diagnostic device 10 starts Doppler mode or M mode in response to the user's instruction from the input interface 28.

[0085] In step S12, in Doppler mode, the Doppler processing unit 18 generates Doppler data as target time series data based on the received beam data string from the receiving unit 16. In M mode, the beam data processing unit 20 generates a received beam data string as target time series data that has undergone various signal processing.

[0086] In step S14, the artifact prediction unit 38 inputs the object time series data (Doppler data or received beam data string) generated in step S12 into the learned artifact prediction learner 32. Then, the artifact prediction unit 38 predicts the type of artifact generated by the object time series data based on the output data of the artifact prediction learner 32 for the object time series data.

[0087] In step S16, the artifact reduction unit 40 performs artifact reduction processing to reduce the type of artifact predicted by the artifact prediction unit 38. Alternatively, step S16 can be bypassed.

[0088] In step S18, the display control unit 24 causes the display 26 to display the corrected ultrasound image that has undergone the artifact reduction processing in step S16, as well as the types of artifacts that have been reduced. Alternatively, if step S16 is bypassed, the display control unit 24 notifies the user of the prediction results (types of artifacts generated in the ultrasound image 52) from the artifact prediction unit 38.

[0089] In step S20, the processor 36 determines whether the Doppler mode or M-mode has ended according to the user's instruction. If the Doppler mode or M-mode continues, the process returns to step S12 and repeats the steps S12 to S20. If the Doppler mode or M-mode has ended, the process ends.

[0090] The above describes the ultrasonic time series data processing apparatus involved in this disclosure. However, the ultrasonic time series data processing apparatus involved in this disclosure is not limited to the above embodiments. Various modifications can be made as long as they do not depart from the main idea.

[0091] For example, in this embodiment, the ultrasound time series data processing device is the ultrasound diagnostic device 10, but the ultrasound time series data processing device is not limited to the ultrasound diagnostic device 10 and may also be other computers. In this case, the learned artifact prediction learner 32 is stored in a memory accessible from the computer that is the ultrasound time series data processing device, and the processor of the computer performs the functions of the artifact prediction unit 38, the artifact reduction unit 40, and the display control unit 24.

Claims

1. An ultrasonic time series data processing device, characterized in that, have: An artifact prediction unit takes object time-series data as input to an artifact prediction learner and predicts the type of artifact generated by the object time-series data based on the output of the artifact prediction learner to the input. The artifact prediction learner learns from time-series data representing the temporal variation of a signal generated by repeated ultrasound transmission and reception at the same location within an examined area, based on reflected waves from the examined area or blood flowing within the examined area, and from time-series data that generates artifacts in ultrasound images based on this time-series data. This time-series data is then used as learning data, along with information indicating the type of artifact. The learning unit predicts and outputs the type of artifact generated by the time-series data based on the input time-series data, which is generated by repeatedly transmitting and receiving ultrasound at the same location within an examined area. An artifact reduction unit performs artifact reduction processing to reduce the type of artifact predicted by the artifact prediction unit; and The display control unit causes the display unit to display an ultrasonic image based on the object's time-series data, which has undergone the artifact reduction processing. When the predicted types of artifacts can be reduced by changing the operating conditions of the ultrasound diagnostic device that generates the object time series data, the artifact reduction unit changes the settings of the ultrasound diagnostic device as an artifact reduction process. When the predicted types of artifacts cannot be reduced by changing the operating conditions of the ultrasound diagnostic device that generates the object time series data, the unit inputs the object time series data into an image generation model that generates an ultrasound image with reduced artifacts based on the time series data that generates the artifacts, and generates an ultrasound image with reduced artifacts as an artifact reduction process.

2. The ultrasonic time series data processing device according to claim 1, characterized in that, The ultrasonic time series data processing device also includes: The notification department will inform the user that the artifact reduction process has been performed.

3. The ultrasonic time series data processing device according to claim 1, characterized in that, The ultrasonic time series data processing device also includes: The time series data generation unit generates the time series data of the object. The artifact prediction unit predicts the types of artifacts generated by the object time series data generated by the time series data generation unit in real time. The artifact reduction unit performs the artifact reduction process in real time based on the artifact type prediction of the artifact prediction unit. The display control unit uses the artifact reduction processing to make the display unit display an ultrasonic image based on the object's time-series data that has undergone the artifact reduction processing in real time.

4. The ultrasonic time series data processing device according to claim 1, characterized in that, The object being inspected and the inspected part being the part that pulsates. The object time series data and the learning time series data are time series data corresponding to the same period in the pulsation cycle of the object and the inspected part.

5. A computer-readable, non-transitory storage medium storing computer-executable commands, characterized in that the commands cause the computer to perform: The artifact prediction step involves inputting object time-series data into an artifact prediction learner, and predicting the types of artifacts generated by the object time-series data based on the output of the learner for that input. The artifact prediction learner is trained using a combination of time-series data representing the temporal variation of a signal generated by repeatedly transmitting and receiving ultrasound waves at the same location within the examined area and based on reflected waves from the examined area or blood flowing within the examined area, and time-series data used to learn artifacts generated in ultrasound images based on this time-series data, and information indicating the type of the artifact. This allows the learner to predict and output the type of artifact generated by the time-series data based on the input time-series data. The object time-series data is generated by repeatedly transmitting and receiving ultrasound waves at the same location within the examined area. The artifact reduction step performs artifact reduction processing to reduce the type of artifact predicted in the artifact prediction step. and The display control step causes the display unit to display an ultrasound image based on the object's time-series data, which has undergone the artifact reduction processing described above. In the artifact reduction step, if the predicted types of artifacts can be reduced by changing the operating conditions of the ultrasound diagnostic device that generates the object time series data, the settings of the ultrasound diagnostic device are changed as an artifact reduction process; if the predicted types of artifacts cannot be reduced by changing the operating conditions of the ultrasound diagnostic device that generates the object time series data, an ultrasound image with reduced artifacts is generated by inputting the object time series data into an image generation model that generates an ultrasound image with reduced artifacts based on the time series data that generates the artifacts, as an artifact reduction process.

Images