KR102248444B1 - Method of generating ultrasonic image and apparatuses performing the same - Google Patents

Abstract

Disclosed are a method of generating an ultrasound image and apparatuses for performing the same. A method of generating an ultrasound image according to an exemplary embodiment includes obtaining an ultrasound image of a blood vessel, and generating an ultra-high resolution ultrasound image in which target particles of the ultrasound image are reflected based on a previously learned neural network algorithm.

Description

Ultrasound image generation method and devices for performing the same BACKGROUND OF THE INVENTION

The following embodiments relate to a method of generating an ultrasound image and apparatuses for performing the same.

It is very important to diagnose circulatory diseases by accurately imaging blood vessel shape and blood flow information.

An ultrasound imaging technique is used to image blood vessels. The ultrasound imaging technique is widely used because it has advantages such as non-invasiveness, economy, and fast image processing.

However, the ultrasound imaging technique is very difficult to measure blood vessels in deep tissues with high resolution due to the disadvantages of the ultrasound imaging technique. Disadvantages of the ultrasound imaging technique may be a trade-off between spatial resolution and depth of penetration, and diffraction limitations.

Recently, the introduction of ultrasonic localization microscopy (ULM) has made it possible to measure blood vessels in deep tissues with high resolution.

First, ULM can detect microbubbles moving along blood vessels in tens of thousands of ultrasound images through a signal decorrelation method. In this case, the microbubbles may be microbubbles formed in blood vessels by an ultrasonic contrast medium. Recently, clinically approved ultrasound contrast agents are commonly used in medical fields, but controversy over the biocompatibility of ultrasound contrast media continues. Therefore, the ultrasound contrast agent has the disadvantage of avoiding the use of the ultrasound contrast agent by patients including pregnant women. In addition, the ultrasonic contrast agent has a disadvantage in that it does not properly function as a target particle (or tracer particle) of blood flow depending on the disappearance rate and concentration of the ultrasonic contrast agent.

Thereafter, the ULM may deconvolution a point spread function (PSF) in the ultrasonic signal of the found micro-bubbles to find the center position of the micro-bubbles.

Finally, the ULM may obtain an ultra-high resolution ultrasound image by reconstructing the center positions of the found microbubbles.

Although the ULM technique can image blood vessels with a resolution smaller than the wavelength, the ULM technique greatly affects the measurement accuracy depending on the concentration of microbubbles injected with an ultrasonic contrast medium.

A method of acquiring a high-resolution blood vessel image is to find as many central locations as possible of the microbubbles flowing inside the blood vessel. In order to find a large number of the center positions of the fine bubbles, the concentration of the fine bubbles must be high, but when the ultrasonic signals overlap due to the high concentration of the fine bubbles, it may be difficult to accurately obtain the center positions of the fine bubbles using a single PSF.

In addition, in an ultrasound image of red blood cells in blood vessels rather than microbubbles by an ultrasound contrast agent, since the signal-to-noise ratio (SNR) of red blood cells is low, it is difficult to obtain accurate location information of red blood cells.

In order to compensate for the above-described problems, various algorithms have been proposed in recent years, but it takes a lot of time to apply the algorithm, and it is difficult to change the optimization setting every time according to the concentration of microbubbles.

The embodiments provide a technology for generating (or converting) an ultrasound image of a blood vessel into an ultrasound image of ultra-high resolution, and generating blood flow information of a blood vessel from the ultra-high resolution ultrasound image to provide an ultra-high resolution ultrasound image and blood flow information. I can.

A method of generating an ultrasound image according to an exemplary embodiment includes obtaining an ultrasound image of a blood vessel, and generating an ultra-high resolution ultrasound image in which target particles of the ultrasound image are reflected based on a previously learned neural network algorithm.

The pre-learned neural network algorithm may be learned using training data obtained by synthesizing a training ultrasound image and information on a training target particle of the training ultrasound image.

The target particle and the learning target particle may be at least one of microbubbles and red blood cells included in the blood vessel or conduit.

The learning data may include first learning data on microbubbles and second learning data on red blood cells.

The generating includes extracting the center position of the target particle using the previously learned neural network algorithm, and generating the ultra-high resolution ultrasound image by reflecting the center position of the target particle in the ultrasound image. can do.

The method includes the steps of generating velocity field information on the internal blood flow of the blood vessel by applying a particle tracking velocity measurement method to a central position of the target particle based on the ultra-high resolution ultrasound image, and the velocity field information based on the velocity field information. It may further include generating flow information about blood flow.

An ultrasound image generating apparatus according to an embodiment includes a communication module and a processor that acquires an ultrasound image of a blood vessel, and generates an ultra-high resolution ultrasound image in which target particles of the ultrasound image are reflected based on a previously learned neural network algorithm. I can.

The pre-learned neural network algorithm may be learned using training data obtained by synthesizing a training ultrasound image and information on a training target particle of the training ultrasound image.

1 is a diagram illustrating a system for generating an ultrasound image according to an exemplary embodiment.
FIG. 2 is a schematic block diagram of the ultrasound image generating apparatus illustrated in FIG. 1.
3 shows an example for explaining the neural network algorithm shown in FIG. 2.
4 shows an example for explaining the operation of the processor shown in FIG. 2.
5 shows an example for explaining a center position of a target particle according to an embodiment.
6 shows an example for explaining a center position tracking accuracy and a center position detection rate of a target particle according to an embodiment.
7 shows an example for explaining an ultrasound image of a branch blood vessel model according to an embodiment.
8 illustrates an example for explaining an ultrasound image and blood flow information for a hydration gel branch blood vessel model according to an exemplary embodiment.
9 illustrates an example for explaining an ultrasound image and blood flow information for an ankle valve of a great saphenous vein according to an exemplary embodiment.
10 is a diagram illustrating an operation of the ultrasound image generating apparatus illustrated in FIG. 1.
11 is a flowchart illustrating an operation of the ultrasound image generating apparatus illustrated in FIG. 1.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the patent application is not limited or limited by these embodiments. It is to be understood that all changes, equivalents, or substitutes to the embodiments are included in the scope of the rights.

The terms used in the examples are used for illustrative purposes only and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof does not preclude in advance.

Terms such as first or second may be used to describe various elements, but the elements should not be limited by terms. The terms are only for the purpose of distinguishing one component from other components, for example, without departing from the scope of rights according to the concept of the embodiment, the first component may be named as the second component, and similarly The second component may also be referred to as a first component.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in the present application. Does not.

In addition, in the description with reference to the accompanying drawings, the same reference numerals are assigned to the same components regardless of the reference numerals, and redundant descriptions thereof will be omitted. In describing the embodiments, when it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the embodiments, the detailed description thereof will be omitted.

A module in the present specification may mean hardware capable of performing functions and operations according to each name described in the present specification, or may mean a computer program code capable of performing a specific function and operation. Or, it may mean an electronic recording medium, for example, a processor or a microprocessor in which a computer program code capable of performing a specific function and operation is mounted.

In other words, the module may mean a functional and/or structural combination of hardware for performing the technical idea of the present invention and/or software for driving the hardware.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the scope of the patent application is not limited or limited by these embodiments. The same reference numerals shown in each drawing indicate the same members.

1 is a diagram illustrating a system for generating an ultrasound image according to an exemplary embodiment.

The ultrasound image generating system 10 may include an ultrasound image providing apparatus 100 and an ultrasound image generating apparatus 300.

The ultrasound image providing apparatus 100 may provide an ultrasound image of a blood vessel to the ultrasound image generating apparatus 300. The ultrasound image may be an ultrasound image with poor resolution.

As an example, the ultrasound image providing apparatus 100 may be a photographing device that actually photographs a blood vessel portion and generates an ultrasonic image of the blood vessel.

As another example, the ultrasound image providing apparatus 100 may be an electronic device of various organizations related to medical care, such as medical facilities and medical institutions having ultrasound images of various blood vessels.

The electronic device may include a database storing various ultrasound images of blood vessels.

For example, the electronic device may be various devices such as a personal computer (PC), a data server, or a portable electronic device. Portable electronic devices include a laptop computer, a mobile phone, a smart phone, a tablet PC, a mobile internet device (MID), a personal digital assistant (PDA), an enterprise digital assistant (EDA). ), digital still camera, digital video camera, portable multimedia player (PMP), personal navigation device or portable navigation device (PND), handheld game console, e-book (e-book), it can be implemented as a smart device (smart device). In this case, the smart device may be implemented as a smart watch or a smart band.

The ultrasound image generating device 300 generates (or converts) an ultrasound image of a blood vessel transmitted from the ultrasound image providing device 100 into an ultra-high resolution ultrasound image, and generates blood flow information of the blood vessel from the ultra-high resolution ultrasound image. Ultra-high resolution ultrasound images and blood flow information may be provided to various devices (eg, display devices, electronic devices, etc.).

Accordingly, the ultrasound image generating apparatus 300 quickly and accurately tracks the center position of the target particle according to a given measurement condition regardless of the presence or absence of an ultrasound contrast agent, and provides the accurate shape of the microvessel and blood flow information, which is hemodynamic information in various clinical conditions. By obtaining, it is possible to generate and provide ultra-high resolution ultrasound images of blood vessels and flow information of blood flow inside the blood vessels.

In addition, the ultrasound image generating device 300 can be applied to improve image quality by converting the image quality of ultrasound clinical data of a medical institution into ultra-high-resolution ultrasound image data by reflecting various conditions, and can be used (or applied) in actual clinical conditions. I can.

In addition, the ultrasound image generating device 300 can be widely used in clinical practice without the patient's reluctance to the ultrasound contrast agent, and can be a cornerstone of fusion research in the field of ultrasound imaging techniques and deep learning-based learning algorithms, and As a result, it can contribute to leading the high value-added medical industry.

As illustrated in FIG. 1, the ultrasound image providing apparatus 100 and the ultrasound image generating apparatus 300 are shown to be independently distinguished, but the present invention is not limited thereto. For example, the ultrasound image providing apparatus 100 and the ultrasound image generating apparatus 300 may be integrated into one device.

FIG. 2 is a schematic block diagram of the ultrasound image generating apparatus illustrated in FIG. 1.

The ultrasound image generating apparatus 300 includes a communication module 310, a memory 330, and a processor 350.

The communication module 310 may transmit the ultrasound image transmitted from the ultrasound image providing apparatus 100 to the processor 350.

The communication module 310 may provide the ultra-high resolution ultrasound image transmitted from the processor 350 to various devices.

The memory 330 may store instructions (or programs) executable by the processor 350. For example, the instructions may include instructions for executing an operation of the processor 350 and/or an operation of each component of the processor 350.

The processor 350 may process data stored in the memory 330. The processor 350 may execute computer-readable code (eg, software) stored in the memory 330 and instructions induced by the processor 350.

The processor 350 may be a data processing device implemented in hardware having a circuit having a physical structure for executing desired operations. For example, desired operations may include code or instructions included in a program.

For example, a data processing device implemented in hardware is a microprocessor, a central processing unit, a processor core, a multi-core processor, and a multiprocessor. , Application-Specific Integrated Circuit (ASIC), Field Programmable Gate Array (FPGA).

The processor 350 may control the overall operation of the ultrasound image generating apparatus 300. For example, the processor 350 may control operations of the components 310 and 330 of the ultrasound image generating apparatus 300.

The processor 350 may acquire an ultrasound image of a blood vessel transmitted from the ultrasound image providing apparatus 100.

The processor 350 may generate an ultra-high resolution ultrasound image in which target particles of the ultrasound image are reflected based on the ultrasound image and the learned neural network algorithm 370. The target particle may be at least one of microbubbles and red blood cells contained in blood vessels or conduits. The micro-bubbles may be micro-bubbles caused by an ultrasonic contrast agent.

For example, the processor 350 may train a neural network algorithm 370 for generating (or converting) an ultrasound image into an ultra-high resolution image. The processor 350 may generate an ultra-high resolution ultrasound image in which target particles of the ultrasound image are reflected by inputting an ultrasound image to the previously learned neural network algorithm 370 as input data.

Also, the processor 350 may generate blood flow information for a blood vessel corresponding to the ultrasound image based on the ultra-high resolution ultrasound image. Blood flow may be blood flow (or blood) inside a blood vessel.

3 shows an example for explaining the neural network algorithm shown in FIG. 2.

The neural network algorithm 370 may be a pre-learned neural network algorithm that is learned (or deep-learning) by learning data obtained by synthesizing information on a learning target particle of a training ultrasound image and a training ultrasound image. The neural network algorithm 370 may be a U-net architecture. The U-net may be a convolutional neural network. The training ultrasound image may be an ultrasound image of various blood vessels previously generated and stored in order to learn a neural network algorithm. The learning target particles are target particles of blood vessels or conduits corresponding to the learning ultrasound image, and may be microbubbles and red blood cells.

The training data may include first training data 710 for microbubbles and second training data 730 for red blood cells.

The first training data 710 may be data obtained by synthesizing a first training ultrasound image of a blood vessel including microbubbles, a speckle signal (or feature) and a center position of the microbubbles of the first training ultrasound image. The speckle signal of the microbubbles of the first training ultrasound image may be a signal reflecting the concentration of the microbubbles, the intensity of the ultrasound signal, and the shape of the first training ultrasound image. The center position of the microbubbles in the first training ultrasound image may be a location at the center of the shape having a high density of microbubbles and an ultrasound signal intensity of the first training ultrasound image.

For example, the first training data 710 may include a first microbubble image and a second microbubble image. The first microbubble image may be a first synthesized ultrasound image in which the center position of the microbubble of the first training ultrasound image is reflected in the first training ultrasound image. The second microbubble image may be a second synthesized ultrasound image in which a speckle signal of microbubbles of the first training ultrasound image is reflected in the first training ultrasound image.

The second training data 730 may be data obtained by synthesizing a second training ultrasound image of a blood vessel including red blood cells and a speckle signal and a center position of red blood cells of the second training ultrasound image. The speckle signal of red blood cells of the second training ultrasound image may be a signal reflecting the concentration, intensity and shape of the red blood cells of the second training ultrasound image. The center position of the red blood cells in the second training ultrasound image may be a location at the center of the second training ultrasound image having a high concentration of red blood cells and an ultrasound signal intensity.

For example, the second learning data 730 may include a first red blood cell image and a second red blood cell image. The first red blood cell image may be a third synthetic ultrasound image in which the center position of the red blood cells of the second training ultrasound image is reflected on the second training ultrasound image. The second red blood cell image may be a fourth synthetic ultrasound image in which a speckle signal of red blood cells of the second training ultrasound image is reflected on the second training ultrasound image.

The first microbubble image and the first red blood cell image may be a binary image (or binary image) in which the training ultrasound image is converted to black, and the training target particles are displayed as white dots with a minimum pixel size in the converted training ultrasound image. .

As described above, although the learning ultrasound image is black and the learning target particles are white, it is not limited thereto. For example, the learning ultrasound image and the learning target particle may be converted and displayed in a color that is distinguished from each other.

The second microbubble image and the second red blood cell image may be images in which speckle signals and gaussian noise of the learning target particles are reflected at the center position of the learning target particles of the binary image. The speckle signal of the learning target particle may be a signal according to a concentration of the learning target particle, an ultrasonic signal intensity, and a shape.

That is, the neural network algorithm 370 is based on the first training data for microbubbles in which the first and second synthetic ultrasound images are paired, and the second training data for red blood cells in which the third and fourth synthetic ultrasound images are paired. It can be a learned U-net architecture. The neural network algorithm 370 may be trained by adjusting the amount of learning until a high accuracy is obtained by checking a tendency of improving accuracy.

4 shows an example for explaining the operation of the processor shown in FIG. 2.

The processor 350 may generate (or convert) an ultrasound image into an ultra-high resolution ultrasound image using the previously learned neural network algorithm 370. The previously learned neural network algorithm may be an ultra-high resolution ultrasound imaging technique (DL-SRU) using deep learning.

First, the processor 350 may track a target particle of an ultrasound image in an ultrasound image using the previously learned neural network algorithm 370.

For example, the processor 350 may track (or detect) target particles of each of the plurality of ultrasound images by inputting a plurality of ultrasound images of a blood vessel to the previously learned neural network algorithm 370. The target particles may be trace particles that need to be tracked in the ultrasound image in order to convert (or generate) the ultrasound image into an ultra-high resolution ultrasound image.

Thereafter, the processor 350 may analyze the target particle and extract the central position of the target particle in the entire area of the target particle. The center position of the target particle is the highest (or greater, higher, or higher) concentration of the target particle and/or the ultrasonic signal intensity in the entire area of the target particle included in the ultrasound image, and is located at the center of the shape of the target particle. It can be a central area. The concentration of target particles may mean the number of target particles observed per unit area. The ultrasonic signal intensity of the target particle may mean an intensity distribution in which values in each pixel are calculated through image processing.

For example, the processor 350 may accurately track and extract (or extract) the center position of the target particle from the speckle signal of the target particle of each of the plurality of ultrasound images. The speckle signal may be a signal in which unique acoustic characteristics according to the concentration, signal strength, and shape of the target particle are reflected. The speckle signal may be a granular signal in the form of an ellipse. The speckle signal for red blood cells may be a signal due to interference of sound waves scattered by red blood cells. The speckle signal for the fine bubbles may be a signal due to interference of sound waves scattered by the fine bubbles.

Finally, the processor 350 may generate an ultra-high resolution ultrasound image by reflecting the center position of the target particle in the ultrasound image.

For example, the processor 350 reflects the center position of the target particle of each of the plurality of ultrasound images on the plurality of ultrasound images, and then superimposes the plurality of ultrasound images reflecting the center position of the target particle to generate an ultra-high resolution ultrasound image. Can be generated.

In addition, the processor 350 may obtain blood flow information about the blood flow (or blood flow) of the ultra-high resolution ultrasound image.

For example, the processor 350 may apply a particle tracking velocity measurement method to the center position of the target particle of the ultra-high resolution ultrasound image to generate velocity field information on the internal blood flow of the blood vessel. The particle tracking velocity measurement method is a particle tracking velocimetry (PTV) algorithm, and may be an algorithm that acquires velocity information of a target particle while tracking a target particle.

The processor 350 may generate blood flow information for a blood vessel based on the velocity field information. The blood flow information may include a velocity vector (eg, direction and size) by blood in each blood vessel region, a pressure distribution, and a wall shear stress in the vessel wall.

5 shows an example for explaining a center position of a target particle according to an embodiment.

5A and 5B show the center position of the micro-bubble tracked by applying the existing ULM technique and the DL-SRU to the ultrasound image of the micro-bubble and the center position of the actual micro-bubble. In this case, the DL-SRU may be a neural network algorithm that is learned from learning data about microbubbles in which the first synthesized ultrasound image and the second synthesized ultrasound image are paired.

When comparing the existing ULM techniques of A and B with the DL-SRU, the DL-SRU of A and B tracks (or measures, detects) the center position of the micro-bubbles more accurately than the existing ULM technique even in the ultrasound image of the superimposed micro-bubbles.

5C shows the center position of red blood cells and the actual center of red blood cells tracked by applying the existing ULM technique and DL-SRU to an ultrasound image of red blood cells. In this case, the DL-SRU may be a neural network algorithm learned as learning data for red blood cells in which the third and fourth synthesized ultrasound images are paired.

When comparing the existing ULM technique of C with the DL-SRU, the DL-SRU of C tracks (or measures, detects) the center position of red blood cells more accurately than the existing ULM technique even if the speckle signal of red blood cells is weak in the actual ultrasound image. Existing ULM techniques cannot accurately track the center position of red blood cells because the speckle signal of red blood cells is weak in actual ultrasound images.

That is, the DL-SRU has superior center position tracking performance for tracking the central position of the target particle than the conventional ULM technique.

6 shows an example for explaining a center position tracking accuracy and a center position detection rate of a target particle according to an embodiment.

6A shows the accuracy of tracking (or measuring) the center position of the target particle of the DL-SRU, indicating how accurately the center position of the target particle of the ultrasound image is tracked (or measured) in an in vitro condition.

6B shows the detection rate of target particles of DL-SRU, indicating how accurately target particles of an ultrasound image are detected (or tracked) under in vitro conditions.

The DL-SRU of FIG. 6A and B were learned while changing the concentration of the learning target particle. Even if the concentration of the learning target particles increased, the accuracy of tracking the center of the DL-SRU and the detection rate of the target particles were high.

Since the experimental results by the DL-SRU and the previous research results are similar, the performance of the DL-SRU of FIG. 6 is similar to that of the previous research.

7 shows an example for explaining an ultrasound image of a branch blood vessel model according to an embodiment.

7A shows a branched blood vessel model obtained in a computer virtual experiment (or computer simulation, in silico).

7B shows an average image of an ultrasound signal for a branch blood vessel model.

7C shows an ultrasound image obtained by reconstructing an ultrasound image of a branch blood vessel model using a conventional ULM technique.

7D shows an ultra-high resolution ultrasound image obtained by reconstructing an ultrasound image of a branch blood vessel model using DL-SRU.

The ultra-high resolution ultrasound image of FIG. 7 may be generated by acquiring a simulation result of a target particle passing through a branch blood vessel model as an ultrasound image, and using the acquired ultrasound image as input data to the DL-SRU. In this case, the ultra-high resolution ultrasound image may be an ultrasound image in which a blood vessel structure is formed by finding a central position of 50,000 or more target particles from 5,000 ultrasound images.

E, F, and G of FIG. 7 denote spatial resolution of an image in which parts of B, C, and D are enlarged.

H and I of FIG. 7 represent the results of analysis by expressing the intensity of the signal according to the dotted lines H-H' and I-I' shown in B, C, and D in a graph.

Comparing B, C, and D of FIG. 7, D has a higher resolution of a blood vessel ultrasound image than B and C generated by conventional imaging techniques. For example, the resolution of the ultrasound image is the spatial resolution of the experimental technique along the dotted lines H-H' and I-I' of the branches divided into two and three branches in B, C, D, E, F, and G. It can be judged by comparing.

8 illustrates an example for explaining an ultrasound image and blood flow information for a hydration gel branch blood vessel model according to an exemplary embodiment.

Fig. 8A shows an ultrasound image of a hydration gel branch vascular model in which the branch blood and the model are made of hydration gel under in vitro conditions. In this case, the ultrasound image may be an ultrasound image of microbubbles of the ultrasound contrast agent.

FIG. 8B shows an average image of an ultrasound signal for a hydration gel branch vascular model.

FIG. 8C shows an ultrasound image obtained by reconstructing an ultrasound image of a hydration gel branch vascular model using a conventional ULM technique.

FIG. 8D shows an ultra-high resolution ultrasound image obtained by reconstructing an ultrasound image of a hydration gel branch vascular model using DL-SRU.

The ultra-high resolution ultrasound image of FIG. 8 may be generated by acquiring an experiment result of a target particle passing through a hydration gel blood vessel model as an ultrasound image, and using the acquired ultrasound image as input data to the DL-SRU. In this case, the ultra-high resolution ultrasound image may be an ultrasound image in which a blood vessel structure is formed by finding a central position of 85,000 or more target particles from 798 ultrasound images.

E of FIG. 8 shows the result of expressing the intensity of a signal converted along the dotted line E-E' shown in B, C, and D as a graph.

8F is the velocity field of the flow inside the vessel model generated (or measured) by applying particle tracking velocimetry (PTV) to the center position of the target particle of the ultra-high resolution ultrasound image reconstructed with DL-SRU. Represents information-based blood flow information.

Comparing B, C, and D of FIG. 8, D generates less noise on the vessel model wall than B and C generated by the conventional imaging technique, and has a higher resolution of a blood vessel ultrasound image than B and C. For example, the resolution of the ultrasound image may be determined by comparing the spatial resolution of each experimental technique along the dotted line E-E' in B, C, and D.

9 illustrates an example for explaining an ultrasound image and blood flow information for an ankle valve of a great saphenous vein according to an exemplary embodiment.

9A shows an ultrasound image of a blood vessel obtained in the vicinity of an ankle valve of a great saphenous vein of a human in vivo. In this case, the ultrasound image may be an ultrasound image of red blood cells without injection of an ultrasound contrast agent.

9B shows an ultrasound image of a blood vessel near an ankle valve of a great saphenous vein, reconstructed using a conventional ULM technique, and a center position of a target particle.

9C shows an ultra-high resolution ultrasound image obtained by reconstructing an ultrasound image of a blood vessel near an ankle valve of a great saphenous vein using DL-SRU and a center position of a target particle.

The DL-SRU of FIG. 9 may be a neural network algorithm learned from an ultrasound image of red blood cells without injecting an ultrasound contrast agent in a blood vessel near an ankle valve of a large saphenous vein of a human lower limb. For example, the DL-SRU may be learned by using a synthetic ultrasound image of red blood cells generated by analyzing speckle signal characteristics of red blood cells.

In the ultra-high resolution ultrasound image of FIG. 9, the experimental result of the target particle passing through the blood vessel near the ankle valve of the great saphenous vein is acquired as an ultrasound image, and the acquired ultrasound image is used as input data to the DL-SRU learned as described above. Can be created by

D1, D2, and D3 of FIG. 9 indicate the accuracy of measuring the center position of the target particle in the enlarged image of positions 1, 2 and 3 indicated in B and C.

9E shows accurate blood flow information near the wall of a blood vessel generated by applying particle tracking velocimetry (PTV) to the center position of a target particle in the ultra-high resolution ultrasound image reconstructed with DL-SRU.

F of FIG. 9 shows an enlarged image of f indicated by E.

Comparing D1, D2, and D3 of FIG. 9, C can accurately find the central position of red blood cells regardless of the location of blood vessels even in various shapes, low intensity, and overlapping conditions of the speckle signal of red blood cells.

10 is a diagram illustrating an operation of the ultrasound image generating apparatus illustrated in FIG. 1, and FIG. 11 is a flowchart illustrating an operation of the ultrasonic image generating apparatus illustrated in FIG. 1.

The processor 350 may learn the neural network algorithm 370 based on the first training data for microbubbles and the second training data for red blood cells (1110).

The processor 350 may acquire an ultrasound image transmitted from the ultrasound image providing apparatus 100 (1130 ).

The processor 350 may generate an ultra-high resolution ultrasound image reflecting target particles of the ultrasound image and blood flow information by inputting an ultrasound image to the previously learned neural network algorithm 370 as input data (1150).

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of these, configuring the processing unit to operate as desired or processed independently or collectively. You can command the device. Software and/or data may be interpreted by a processing device or, to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodyed in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

As described above, although the embodiments have been described by the limited drawings, a person of ordinary skill in the art can apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order from the described method, and/or components such as systems, structures, devices, circuits, etc. described are combined or combined in a form different from the described method, or other components Alternatively, even if substituted or substituted by an equivalent, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and claims and equivalents fall within the scope of the following claims.

Claims (8)

Obtaining an ultrasound image of a blood vessel;
Generating an ultra-high resolution ultrasound image reflecting the target particles of the ultrasound image based on a previously learned neural network algorithm;
Generating velocity field information on the internal blood flow of the blood vessel by applying a particle tracking velocity measurement method to the center position of the target particle based on the ultra-high resolution ultrasound image; And
Generating flow information for the blood flow based on the velocity field information
Including,
The pre-learned neural network algorithm is a method of generating an ultrasound image learned by using training data obtained by synthesizing a training ultrasound image and information on a target particle of the training ultrasound image.
The method of claim 1,
The method of generating an ultrasound image, wherein the target particle and the learning target particle are at least one of microbubbles and red blood cells included in the blood vessel or conduit.
The method of claim 1,
The training data,
First learning data for fine bubbles; And
Second learning data for red blood cells
Ultrasound image generation method comprising a.
The method of claim 3,
The first training data is data obtained by synthesizing a first training ultrasound image of a blood vessel including the micro-bubbles, a speckle signal and a center position of the micro-bubbles,
The second learning data is data obtained by synthesizing a second learning ultrasound image of a blood vessel including the red blood cells, a speckle signal of the red blood cells, and a center position of the red blood cells.
The method of claim 1,
The generating step,
Extracting the center position of the target particle using the previously learned neural network algorithm; And
And generating the ultra-high resolution ultrasound image by reflecting the center position of the target particle on the ultrasound image.
delete Communication module; And
Acquires an ultrasound image of a blood vessel, generates an ultra-high resolution ultrasound image reflecting the target particle of the ultrasound image based on a previously learned neural network algorithm, and tracks the particle at the center position of the target particle based on the ultra-high resolution ultrasound image Processor for generating velocity field information on the internal blood flow of the blood vessel by applying a velocity measurement method, and generating flow information on the blood flow based on the velocity field information
Including,
The pre-learned neural network algorithm is a device for generating an ultrasound image that is learned using training data obtained by synthesizing a training ultrasound image and information on a training target particle of the training ultrasound image.
The method of claim 7,
The processor,
An ultrasound image generating apparatus performing the method of generating an ultrasound image according to any one of claims 2 to 5.

Images