KR102655333B1 - Method and apparatus for multi-variable quantitative imaging using ultrasound data - Google Patents

Abstract

A method of operating an imaging device operated by at least one processor, comprising the steps of receiving ultrasound data obtained from a tissue, extracting a feature corresponding to a target variable selected from a plurality of variables from the ultrasound data, and extracting the feature from the ultrasound data. It includes the step of restoring and outputting a quantitative image of the target variable.

Description

Multivariate quantitative imaging method and device using ultrasound data {METHOD AND APPARATUS FOR MULTI-VARIABLE QUANTITATIVE IMAGING USING ULTRASOUND DATA}

The present invention relates to ultrasound imaging.

Cancer is difficult to detect early, requiring periodic diagnosis, and the size and characteristics of lesions must be continuously monitored. Representative imaging equipment for this include X-ray, magnetic resonance imaging (MRI), computed tomography (CT), and ultrasound. While X-rays, MRI, and CT have the disadvantages of risk of radiation exposure, long measurement time, and high cost, ultrasound imaging equipment is safe, relatively inexpensive, and provides real-time images, allowing users to monitor the lesion in real time and obtain the desired image. You can get it.

Currently, the most commercialized ultrasound imaging equipment is the B-mode (Brightness mode) imaging system. The B-mode imaging method is a method of determining the location and size of an object through the time and intensity at which ultrasonic waves are reflected from the surface of the object and returned. Because it locates the lesion in real time, the user can efficiently obtain the desired image while monitoring the lesion in real time, and it is safe and relatively inexpensive, making it highly accessible. However, it has the disadvantage of not maintaining consistent image quality depending on the user's skill level and not being able to image quantitative characteristics. In other words, because the B-mode technique provides only morphological information of the tissue, sensitivity and specificity may be low in differential diagnosis for distinguishing between benign and malignant tumors based on histological characteristics.

Pathological changes in tissue cause structural changes in cells, and this involves imaging changes in ultrasound properties in the tissue. Representative examples include elastography and ultrasound computed tomography (USCT). there is. Elastography techniques can quantitatively image tissue elasticity and stiffness, but require additional dedicated devices and consume a lot of energy. Therefore, elastography techniques can only be applied to expensive ultrasound equipment and have a low frame rate, making them unsuitable for imaging dynamically moving tissues. Ultrasound tomography can obtain high-resolution quantitative images, but the ultrasound sensor must surround the object, so it is limited to mammography and has limitations in measuring various organs. In addition, the ultrasound tomography technique takes minutes for imaging, so real-time movement cannot be seen, and the size of the system is so large that it is impossible to move.

The present disclosure provides a method and device for imaging quantitative characteristics of various variables of a tissue by entering ultrasound signals beamformed with different beam patterns into the tissue and using multi-pattern ultrasound data reflected from the tissue. It is provided.

The present disclosure uses multi-pattern ultrasound data acquired from tissue through a single ultrasound probe to measure attenuation coefficient (AC), speed of sound (SoS), and effective scatterer concentration (ESC). , provides a neural network that complexly restores quantitative characteristics of variables including effective scatterer diameter (ESD).

The present disclosure uses a conditional encoder that variably encodes ultrasound data by changing network parameters to optimally extract the target variable from ultrasound data to determine the quantitative characteristics of each variable to be restored. The aim is to provide a method and device to improve image restoration performance for variables.

According to one embodiment, a method of operating an imaging device operated by at least one processor, comprising: receiving ultrasound data obtained from a tissue, and selecting a feature corresponding to a target variable selected from a plurality of variables from the ultrasound data. Extracting and restoring the features and outputting a quantitative image of the target variable.

The ultrasound data may be multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue using different beam patterns are reflected from the tissue and returned.

The multiple variables include two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD). It can be included.

The step of outputting a quantitative image of the target variable may include extracting features corresponding to the target variable from the ultrasound data by changing network parameters of the neural network according to the target variable selected from among the plurality of variables.

The neural network can perform conditional encoding to normalize to a value selected according to the target variable.

In the step of outputting a quantitative image of the target variable, the quantitative image of the target variable may be output through a decoder of the neural network that restores the features output from the encoder of the neural network through the same path.

The decoder may be composed of a network that restores the features of each variable while preserving structural similarity in a path to restore the features of different variables.

An imaging device according to another embodiment includes a memory for loading a computer program, and executing the computer program to extract a feature corresponding to a target variable selected from a plurality of variables from ultrasound data obtained from a tissue, and the feature and a processor that restores and outputs a quantitative image of the target variable.

The ultrasound data may be multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue using different beam patterns are reflected from the tissue and returned.

The multiple variables include two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD). It can be included.

The processor may output a quantitative image corresponding to a target variable selected from among a plurality of variables from the ultrasound data through a neural network composed of an encoder and a decoder. The encoder may be configured as a network that changes network parameters according to the target variable selected from among the plurality of variables and extracts features corresponding to the target variable from the ultrasound data. The decoder may be configured as a network that restores the features output from the encoder through the same path.

The neural network can perform conditional encoding to normalize to a value selected according to the target variable.

The decoder may be composed of a network that restores the features of each variable while preserving structural similarity in a path to restore the features of different variables.

According to another embodiment, a computer program is stored in a computer-readable storage medium and includes instructions executed by a processor, wherein ultrasound data obtained from a tissue is input and a target is selected among a plurality of variables that can be restored from the ultrasound data. When a variable is selected, an encoder changes network parameters according to the selected target variable to extract features corresponding to the target variable from the ultrasound data, and restores the features output from the encoder to produce a quantitative image of the target variable. Contains instructions that execute the output decoder.

The ultrasound data may be multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue using different beam patterns are reflected from the tissue and returned.

The multiple variables include two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD). It can be included.

The encoder may include a conditional encoding block that normalizes to a value selected according to the target variable.

The decoder may be composed of a network that restores the features of each variable while preserving structural similarity in a path to restore the features of different variables.

According to an embodiment, a variety of quantitative variables can be imaged in real time from tissue ultrasound data through a single neural network.

According to an embodiment, a high-quality restored image can be generated through information shared in the process of restoring a quantitative image of variables having structural similarity.

According to the embodiment, since the ultrasound probe and imaging device for B-mode (brightness mode) imaging can be used as is, imaging is simple and various organs that can be measured with existing ultrasound imaging equipment can be measured.

According to an embodiment, a high-quality quantitative image can be restored regardless of the user's skill level using a single ultrasound probe.

Figure 1 is a diagram conceptually explaining a multivariate quantitative imaging device using ultrasound data.
Figure 2 is a diagram illustrating a neural network that extracts quantitative characteristics of tissue from multi-pattern ultrasound data.
Figure 3 is a diagram explaining conditional normalization.
Figure 4 is an example of a neural network encoder that restores multivariate quantitative images.
Figure 5 is an example of a neural network decoder that restores multivariate quantitative images.
Figure 6 is a flowchart of a multivariate quantitative imaging method using ultrasound data according to an embodiment.
Figure 7 is an example of a phantom for multivariate quantitative simulation.
Figure 8 is a diagram showing the results of multivariate quantitative imaging of a breast-mimicking phantom.
Figure 9 is a diagram showing ex-vivo multivariate quantitative imaging results.
Figure 10 is a configuration diagram of a computing device according to one embodiment.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part is said to “include” a certain element, this means that it may further include other elements rather than excluding other elements, unless specifically stated to the contrary. In addition, terms such as "... unit", "... unit", and "module" used in the specification refer to a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. there is.

The neural network of the present invention is an artificial intelligence model that learns at least one task, and can be implemented as software/programs running on a computing device. The program is stored on a non-transitory storage media and includes instructions written to execute the operations of the present invention by a processor. Programs can be downloaded over the network or sold in product form.

Figure 1 is a diagram conceptually explaining a multivariate quantitative imaging device using ultrasound data, Figure 2 is a diagram illustrating a neural network that extracts quantitative characteristics of tissue from multi-pattern ultrasound data, and Figure 3 is a diagram showing conditional normalization. This is a diagram explaining conditional normalization.

Referring to FIG. 1, a multivariate quantitative imaging device (simply referred to as 'imaging device') 100 is a computing device operated by at least one processor, and is a computing device operating by at least one processor. Multi-pattern ultrasound data is input. The imaging device 100 is equipped with a computer program for the operations described in this disclosure, and the computer program is executed by a processor.

The imaging device 100 may complexly generate multi-variable quantitative images of the tissue using a neural network 200 that extracts quantitative characteristics of the tissue from multi-pattern ultrasound data. Here, the imaging device 100 measures quantitative variables of the tissue, such as attenuation coefficient (AC), speed of sound (SoS), effective scatterer concentration (ESC), which represents the density distribution within the tissue, and tissue The effective scatterer diameter (ESD), which indicates the size of inner cells, etc., can be restored. The neural network 200 may be called SQI-Net.

Meanwhile, the attenuation coefficient, sound speed, scatterer density, and scatterer size are each parameters known as biomarkers for lesion extraction and are closely related to the biomechanical characteristics of the tissue. Therefore, the more variables used, the more comprehensive analysis of lesions can be performed, thereby increasing diagnostic sensitivity and specificity. However, because each variable has a distinct effect on ultrasound propagation, it is not easy to simultaneously restore multiple variables from ultrasound data.

To solve this problem, the neural network 200 performs conditional encoding to variably extract quantitative features from ultrasound data according to the target variable to be restored. Through this, the neural network 200 can complexly generate multi-variable quantitative images of the tissue. Here, conditional encoding can improve image restoration performance for the variable by conditionally changing the network parameters of the encoding path according to the selected variable so that the quantitative characteristics of the variable to be restored from ultrasound data are optimally extracted.

While the neural network 200 restores a quantitative image, the characteristics of independent variables provide complementary information. Therefore, even if the neural network 200 individually restores the quantitative image of each variable, the independent variables provide complementary information, allowing the quantitative value to be restored more accurately.

The features of each variable extracted from the conditional encoder are restored through the same decoding path. Accordingly, each quantitative image is individually restored according to variables, but quantitative images may have structural similarities. Accordingly, since the neural network 200 restores multivariate quantitative images while maintaining structural similarity, the imaging device 100 can generate high-quality restored images.

Multi-pattern ultrasound data used to reconstruct multivariate quantitative images of tissue are obtained from the ultrasound probe 10. The ultrasound probe 10 is a probe that radiates ultrasound signals and can obtain ultrasound data reflected from tissue. Ultrasound signals radiated to tissue may be plane waves. The ultrasonic probe 10 has N (eg, 128) ultrasonic sensors arranged, and may be of various types depending on the arrangement shape. Sensors can be implemented with piezoelectrical elements. Additionally, the ultrasonic probe 10 may be a phased array probe that generates an ultrasonic signal by applying an electrical signal to each piezoelectric element at regular time intervals. For reference, the ultrasound probe 10 may be a typical B-mode imaging probe. In this case, the imaging device 100 receives RF data used for B-mode imaging, which is referred to as ultrasound data in the description.

The ultrasound probe 10 can make ultrasound signals of different beam patterns (Tx pattern #1 to #k) incident on the tissue and acquire RF (Radio Frequency) data reflected from the tissue and returned. In the description, RF data obtained by the ultrasonic probe 10 for an ultrasonic signal radiated with a specific beam pattern is referred to as ultrasonic data of a specific beam pattern. Ultrasound data obtained from a plurality of beam patterns can be bundled and simply referred to as multi-pattern ultrasound data or beamformed ultrasound data. Multi-pattern ultrasound data is, for example, ultrasound data obtained from seven different beam patterns (θ _1:: θ ₇ ), and the different beam patterns have an incident angle of, for example, -15°, -10°, It can be set to -5°, 0°, 5°, 10°, and 15°.

Meanwhile, the ultrasonic data obtained from the ultrasonic probe 10 includes delay time information for receiving the reflected ultrasonic signal for each sensor of the ultrasonic probe 10. Therefore, ultrasound data can be expressed as an image representing delay time information for each sensor. For convenience, ultrasound data expressed as an image may be called an ultrasound data image, and the ultrasound data image may be used as an input image of the neural network 200.

Referring to FIG. 2, the neural network 200 is an artificial intelligence model capable of learning at least one task, and may be implemented as software/program running on a computing device. The neural network 200 has an encoder 210 that encodes multi-pattern ultrasound data U according to the target variable q to be restored, and generates a high-resolution quantitative image I _q of the target variable by increasing the resolution of the encoded low-resolution feature map. It includes a decoder 230 that does.

The encoder 210 may receive multi-pattern ultrasound data U expressed as multi-pattern ultrasound data images 300. The decoder 230 outputs a high-resolution quantitative image I _q corresponding to the target variable, including an attenuation coefficient image (400-1), a sound velocity image (400-2), a scatterer density image (400-3), and a scatterer density image (400-3). In addition to the size image 400-4, images in which a plurality of variables are combined according to the target variable can be output.

The encoder 210 receives ultrasound data images indicating delay times for each sensor of the ultrasound data obtained from the ultrasound probe 10, and extracts quantitative features included in the ultrasound data images. At this time, the encoder 210 may be a conditional encoder that variably extracts quantitative features of input images depending on the target variable to be restored.

The encoder 210 changes the network parameters of the encoding path according to the target variable q, and performs conditional encoding to optimally extract quantitative features of the variable to be restored.

행렬에서 선택된 값인

및

으로, 정규화된 값

을 변환(스케일링 및 이동)할 수 있다.

는 목적 변수 q에 대한 스케일링(scaling) 값이고,

는 목적 변수 q에 대한 이동(shift) 값이다. 이를 통해, 인코더(210)는 목적 변수에 따라 입력 이미지들의 정량적 특징을 가변적으로 추출할 수 있다. 인코더(210)는, 예를 들면, 조건부 인스턴스 정규화(conditional instance normalization)을 수행해서, 목적 변수에 대응하는 정량적 특징을 추출할 수 있다. 목적 변수 q는 조직에서 추출하고자 하는 적어도 하나의 정량적 변수를 나타낸다.Referring to FIG. 3, the encoder 210, according to the target variable q,

The value selected from the matrix

and

, the normalized value

You can transform (scale and move).

is the scaling value for the objective variable q,

is the shift value for the objective variable q. Through this, the encoder 210 can variably extract quantitative features of input images depending on the target variable. The encoder 210 may, for example, perform conditional instance normalization to extract quantitative features corresponding to the target variable. The objective variable q represents at least one quantitative variable to be extracted from the organization.

Referring again to FIG. 2, the decoder 230 restores the features extracted from the encoder 210 through the same path. The decoder 230 is trained to restore the features of each variable extracted from the encoder 210 while preserving structural similarity in the process of restoring them through the same decoding path. Therefore, even if the decoder 230 restores the characteristics of a single target variable, it outputs an image that has structural similarity to other variables, and thus can generate a quantitative image with much higher resolution than a conventional quantitative imaging method.

The objective function G ^* used for learning of the neural network 200 can be defined as a loss function as shown in Equation 1. The neural network 200 can learn to minimize restoration loss using the objective function G ^* .

In Equation 1, the objective function G ^* is a function that minimizes the difference between the output G(U,q) inferred from the input U and the objective variable q and the ground truth Y _q for the objective variable q. The objective function G ^* can be L ₂ regularized (λ = 10 ⁶ ) to avoid overfitting.

The target variable q can be changed for each iteration so that the neural network 200 can learn variables at a similar level.

Figure 4 is an example of an encoder of a neural network that restores multivariate quantitative images, and Figure 5 is an example of a decoder of a neural network that restores multivariate quantitative images.

Referring to FIG. 4, the encoder 210 of the neural network 200 receives multi-pattern ultrasound data images 300. Each ultrasound data image U _i ∈ ^{R 128} For example, it can be expressed as 3018).

The encoder 210 can configure an encoding path with various network models/network blocks.

The encoder 210 convolutionally filters and integrates each of the multi-pattern ultrasound data images U, constructs an encoding network that compresses feature values according to the objective variable q, and generates a feature map f extracted from the input U through the encoding network. Output _q (U). The encoding network can be configured in various ways.

The encoder 210 performs convolutional filtering (2d-Convolution, 3×3 Kernel size) and activation (e.g., ReLU activation function) block 211 and down-sampling (e.g., ReLU) for each channel corresponding to each input image. For example, the 1x2 stride) block 212 can be configured independently.

행렬에서 선택된 값인

및

을 스케일링(scaling) 팩터 및 이동(shift) 팩터로 입력받고, 목적 변수에 따라 특징값을 변환할 수 있다. 조건부 인코딩 블록(213)은 예를 들면, 컨볼루션 필터링(2d-Convolution, 3×3 Kernel size) 블록(214), 조건부 인스턴스 정규화(conditional instance normalization) 블록(215), 그리고 활성화 블록(ReLU)(216)을 포함할 수 있다. The filtered feature values in each channel are integrated on a channel basis and then input to the conditional encoding block 213. The conditional encoding block 213 is configured according to the target variable q,

The value selected from the matrix

and

can be input as a scaling factor and a shift factor, and the feature value can be converted according to the target variable. The conditional encoding block 213 includes, for example, a convolution filtering (2d-Convolution, 3×3 Kernel size) block 214, a conditional instance normalization block 215, and an activation block (ReLU) ( 216) may be included.

Subsequently, the encoder 210 uses subsequent conditional encoding blocks 213 to extract feature values in the time domain and space domain for quantitative variables included in the multi-pattern ultrasound data.

For example, the encoder 210 may repeatedly (e.g., three times) concatenate a down-sampling block (e.g., 1x2 stride) 217 and a subsequent conditional encoding block 213 to obtain features in the time domain. You can construct a time domain encoding block to extract the values. Through an encoding path that repeatedly passes through the down-sampling block 217 and the conditional encoding block 213, correlations in the time domain of quantitative variables included in multi-pattern ultrasound data can be extracted. Passing through the time domain encoding block, 7 pieces of ultrasound data R ^128X3018 can be encoded into 128 feature maps R ^128X186X128 of size (128X186).

In order to extract features in the spatial domain, the encoder 210 connects the down-sampling block (e.g., 2x2 stride) 218 and the subsequent conditional encoding block 213 repeatedly (e.g., twice). , it is possible to construct a spatial domain encoding block that extracts feature values in the spatial domain. Passing through the spatial domain encoding block, the extracted feature map of R ^32X47X128 may be encoded into a feature map f _q (U) of R ¹⁶

The encoder 210 outputs a feature map f _q (U) of multi-pattern ultrasound data U extracted according to the objective variable q. The output f _q (U) is input to the decoder 230.

Referring to FIG. 5, the decoder 230 of the neural network 200 receives the output f _q (U) of the encoder 210 and generates a high-resolution quantitative image I _q corresponding to the target variable q. For example, the decoder 230 can generate a high-resolution quantitative image I _q ∈R ^128X128 from f _q (U)∈R ^16X16X512 .

The decoder 230 can configure a decoding network with various network models/network blocks. For example, the decoder 230 can generate a high-resolution quantitative image I _q using an upsampling method. Alternatively, the decoder 230 can generate a high-resolution quantitative image I _q using parallel multi-resolution subnetworks based on the recently known High-Resolution Network (HRNet). .

The decoder 230, which consists of parallel multi-resolution subnetworks, sequentially performs multi-resolution convolution starting from the low-resolution subnetwork, increasing the image resolution and finally generating a high-resolution quantitative image I _q. can do. The output layer of the decoder 230 can merge into the highest resolution representation and generate a high-resolution quantitative image I _q synthesized through 1x1 convolution. Meanwhile, the unit subnetwork can be implemented with, for example, four residual blocks, through which learning stability can be guaranteed.

Multi-resolution convolution can preserve the low-resolution feature map f _q (U) while minimizing information loss while generating high-resolution quantitative images by integrating various quantitative profiles. Multi-resolution fusion is implemented in all nodes of subnetworks to exchange information through multi-resolution representation, which plays an important role in extracting high-resolution quantitative images I _q .

Figure 6 is a flowchart of a multivariate quantitative imaging method using ultrasound data according to an embodiment.

Referring to FIG. 6, the imaging device 100 receives multi-pattern ultrasound data obtained from tissue (S110). Multi-pattern ultrasound data is RF data obtained from ultrasound signals radiated with different beam patterns, and may be referred to as beamformed ultrasound data.

The imaging device 100 changes network parameters according to the target variable to be restored among a plurality of variables, extracts features corresponding to the selected target variable from multi-pattern ultrasound data, and restores the features corresponding to the target variable to provide high resolution. A quantitative image is output (S120). To this end, the imaging device 100 may use a neural network 200 that has been trained to receive multi-pattern ultrasound data and output a quantitative image corresponding to the selected variable. The neural network 200 includes a conditional encoder 210 that changes network parameters according to the target variable to be restored among a plurality of variables and extracts features corresponding to the selected target variable from multi-pattern ultrasound data, and a conditional encoder 210. It may include a decoder 230 that restores the features of the target variable extracted from through the same path.

A target variable to be restored may be selected from a number of variables including attenuation coefficient, speed of sound, scatterer density, scatterer size, etc. The target variable may be a single variable or may be multiple variables. For example, when the target variable is selected as the attenuation coefficient, the imaging device 100 may extract quantitative features of the attenuation coefficient from the input multi-pattern ultrasound data and output a quantitative image expressing the attenuation coefficient. When the target variables are selected as the attenuation coefficient and sound speed, the imaging device 100 can extract quantitative features of the attenuation coefficient and sound speed from the input multi-pattern ultrasound data and output a quantitative image in which the attenuation coefficient and sound speed are expressed. there is.

Figure 7 is an example of a phantom for multivariate quantitative simulation.

Referring to FIG. 7, the learning data of the neural network 200 may consist of ultrasound data obtained from various human environments and may be collected using an ultrasound simulation tool. It can be confirmed that it is possible to extract quantitative information about tissue through ultrasound data measured through sophisticated simulation.

(a) is the simulated sound velocity distribution, (b) is the simulated attenuation coefficient distribution, and (c) is the simulated density distribution.

Simulation phantoms representing organs and lesions are modeled with the sound velocity distribution (a), attenuation coefficient distribution (b), and density distribution (c) to mimic the human body while remaining simple and general. It can be.

For example, in a simulation, the Region of Interest (RoI) may be set to 45mm x 45mm and ellipses with a radius of 2 to 30 mm may be randomly placed. The background and ellipse can have sound speeds ranging from 1400 to 1700 m/s, attenuation coefficients ranging from 0 to 1.5 dB/cm/MHz, and density values ranging from 0.9 to 1.1 kg/m ³ . Square speckles with sizes ranging from 25 to 150 μm can be placed at concentrations of 0 to 10/wavelength ² to express scatterer density and scatterer size.

Figure 8 is a diagram showing the results of multivariate quantitative imaging of a breast-mimicking phantom, and Figure 9 is a diagram showing the results of ex-vivo multivariate quantitative imaging.

Referring to FIG. 8, in a breast-mimicking phantom containing a cyst mass and cancer tissue, the B-mode imaging technique, a conventional image restoration technique, cannot distinguish cyst/cancer tissue (bottom left).

On the other hand, the imaging device 100 can restore the attenuation coefficient (AC), speed of sound (SoS), scatterer density (ESC), and scatterer size (ESD), and through this, it is easy to determine that the observed tissue is a cancer tissue. It is possible to distinguish and restore quantitative variables with a high accuracy of over 95%. For example, the imaging device 100 may detect an attenuation coefficient (AC) that is greater than normal, a speed of sound (SoS) that is faster than normal, a scatterer density (ESC) that is higher than normal, or a scatterer size (ESD) that is higher than normal. If is smaller than normal, it can be identified as cancerous tissue. The imaging device 100 may determine cancer tissue by combining multivariate results to increase accuracy.

Referring to FIG. 9, benign (a), malignant (b), and cancer-like substances (c) are inserted into the skeletal muscle of a cow, and quantitative variables are restored through the imaging device 100 to differentially diagnose cancer. The clinical significance of was verified.

Looking at the B-mode image (second row) acquired using the B-mode imaging technique, a conventional image restoration technique, it can be seen that differential diagnosis of the inserted tissue is very difficult. On the other hand, the imaging device 100 can restore attenuation coefficient (AC), speed of sound (SoS), scatterer density (ESC), and scatterer size (ESD), which are closely related to the biomechanical properties of tissue. Therefore, by analyzing the speed of sound (SoS) and tissue density (ESC) restored through the imaging device 100, cysts and cancer tissues can be easily distinguished. Additionally, the attenuation coefficient (AC) and tissue cell size (ESD) restored through the imaging device 100 can be analyzed to determine whether the tissue is positive/negative.

Figure 10 is a configuration diagram of a computing device according to one embodiment.

Referring to FIG. 10, the imaging device 100 may be a computing device 500 operated by at least one processor and connected to the ultrasonic probe 10 or a device that provides data acquired from the ultrasonic probe 10. do.

The computing device 500 includes one or more processors 510, a memory 530 that loads a program executed by the processor 510, a storage 550 that stores programs and various data, a communication interface 570, and these. It may include a connecting bus 590. In addition, the computing device 500 may further include various components. When loaded into the memory 530, the program may include instructions that cause the processor 510 to perform methods/operations according to various embodiments of the present disclosure. That is, the processor 510 can perform methods/operations according to various embodiments of the present disclosure by executing instructions. Instructions are a series of computer-readable instructions grouped by function and are a component of a computer program and are executed by a processor.

The processor 510 controls the overall operation of each component of the computing device 500. The processor 510 includes at least one of a Central Processing Unit (CPU), Micro Processor Unit (MPU), Micro Controller Unit (MCU), Graphic Processing Unit (GPU), or any type of processor well known in the art of the present disclosure. It can be configured to include. Additionally, the processor 510 may perform operations on at least one application or program to execute methods/operations according to various embodiments of the present disclosure.

The memory 530 stores various data, commands and/or information. Memory 530 may load one or more programs from storage 550 to execute methods/operations according to various embodiments of the present disclosure. The memory 530 may be implemented as a volatile memory such as RAM, but the technical scope of the present disclosure is not limited thereto.

Storage 550 may store programs non-temporarily. The storage 550 is a non-volatile memory such as Read Only Memory (ROM), Erasable Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), flash memory, a hard disk, a removable disk, or a device well known in the art to which this disclosure pertains. It may be configured to include any known type of computer-readable recording medium.

The communication interface 570 supports wired and wireless communication of the computing device 500. To this end, the communication interface 570 may be configured to include a communication module well known in the technical field of the present disclosure.

Bus 590 provides communication functions between components of computing device 500. The bus 590 may be implemented as various types of buses, such as an address bus, a data bus, and a control bus.

As such, according to the embodiment, a variety of quantitative variables can be imaged in real time from tissue ultrasound data through a single neural network.

According to an embodiment, a high-quality restored image can be generated through information shared in the process of restoring a quantitative image of variables having structural similarity.

According to the embodiment, since the ultrasound probe and imaging device for B-mode (brightness mode) imaging can be used as is, imaging is simple and various organs that can be measured with existing ultrasound imaging equipment can be measured.

According to an embodiment, a high-quality quantitative image can be restored regardless of the user's skill level using a single ultrasound probe.

The embodiments of the present invention described above are not only implemented through devices and methods, but can also be implemented through programs that implement functions corresponding to the configurations of the embodiments of the present invention or recording media on which the programs are recorded.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims are also possible. It falls within the scope of rights.

Claims (18)

1. A method of operating an imaging device operated by at least one processor, comprising:
A step of receiving ultrasound data obtained from the tissue, and
From the ultrasound data, changing network parameters of a neural network according to a target variable selected from among a plurality of variables to extract features corresponding to the target variable, and restoring the features to output a quantitative image of the target variable.
An operation method comprising: In paragraph 1:
The ultrasound data is multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue with different beam patterns are reflected from the tissue and returned. In paragraph 1:
The above multiple variables are
An operation method comprising two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD) . delete In paragraph 1:
The neural network performs conditional encoding to normalize to a value selected according to the target variable. In paragraph 1:
The step of outputting a quantitative image of the target variable is
An operating method for outputting a quantitative image of the target variable through a decoder of the neural network that restores the features output from the encoder of the neural network through the same path. In paragraph 6:
The decoder is comprised of a network that restores the features of each variable while preserving structural similarity in a path for restoring the features of different variables. memory that loads computer programs, and
Executing the computer program, the network parameters of the neural network are changed according to the target variable selected from among a plurality of variables from the ultrasound data obtained from the tissue to extract features corresponding to the target variable, and the features are restored to achieve the target variable. Processor that outputs quantitative images of variables
Including, an imaging device. In paragraph 8:
The ultrasound data is multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue with different beam patterns are reflected from the tissue and returned. In paragraph 8:
The above multiple variables are
An imaging device comprising two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD) . In paragraph 8:
The processor is
Outputting a quantitative image corresponding to a target variable selected from among a plurality of variables from the ultrasound data through a neural network composed of an encoder and a decoder,
The encoder consists of a network that changes network parameters according to the target variable selected from among the plurality of variables and extracts features corresponding to the target variable from the ultrasound data,
The decoder is configured as a network that restores the features output from the encoder through the same path. In paragraph 11:
The neural network performs conditional encoding to normalize to a value selected according to the target variable. In paragraph 11:
The decoder is comprised of a network that restores the features of each variable while preserving structural similarity in a path to restore the features of different variables. A computer program stored in a computer-readable storage medium and including instructions executed by a processor,
When ultrasound data obtained from a tissue is input and a target variable is selected from among a plurality of variables that can be restored from the ultrasound data, network parameters are changed according to the selected target variable, and features corresponding to the target variable are selected from the ultrasound data. an encoder that extracts, and
A computer program comprising instructions for executing a decoder that restores the features output from the encoder and outputs a quantitative image of the target variable. In paragraph 14:
The ultrasound data is multi-pattern RF (Radio Frequency) data in which ultrasound signals incident on the tissue with different beam patterns are reflected from the tissue and returned, a computer program. In paragraph 14:
The above multiple variables are
A computer program that includes two or more of Attenuation Coefficient (AC), Speed of Sound (SoS), Effective Scatterer Concentration (ESC), and Effective Scatterer Diameter (ESD) . In paragraph 14:
A computer program, wherein the encoder includes a conditional encoding block that normalizes to a value selected according to the target variable. In paragraph 14:
The decoder is a computer program that consists of a network that restores the features of each variable while preserving structural similarity in a path to restore the features of different variables.