WO2021120065A1

WO2021120065A1 - Automatic measurement method and ultrasonic imaging system for anatomical structure

Info

Publication number: WO2021120065A1
Application number: PCT/CN2019/126388
Authority: WO
Inventors: 邹耀贤; 林穆清; 王泽兵
Original assignee: 深圳迈瑞生物医疗电子股份有限公司
Priority date: 2019-12-18
Filing date: 2019-12-18
Publication date: 2021-06-24
Also published as: CN112998755A

Abstract

An automatic measurement method and an ultrasonic testing system for an anatomical structure. The automatic measurement method comprises: an image acquisition step (11), involving: acquiring an ultrasound image, wherein the ultrasound image is associated with at least one anatomical structure of a living body tissue, and the at least one anatomical structure has at least one measurement item; a measurement item acquisition step (12), involving: acquiring a measurement item to be measured; an image recognition step (13), involving: recognizing, from the ultrasound image, a recognition image containing a measurement target corresponding to said measurement item; and an automatic measurement step (14), involving: measuring the measurement target in the recognition image. An ultrasound image containing a measurement target corresponding to a measurement item to be measured is recognized according to said measurement item needing to be measured, and the measurement target in the ultrasound image is measured automatically, such that the efficiency of measurement of an anatomical structure is improved.

Description

Automatic measurement method of anatomical structure and ultrasound imaging system

Manual

Technical field

This application relates to the field of medical equipment, and more specifically to an automatic measurement method of anatomical structure and an ultrasound imaging system.

Background technique

Ultrasound measurement is a common method to obtain the size of tissues or lesions. In order to improve the measurement effect, many ultrasound manufacturers have integrated automatic measurement algorithms. For example, for obstetric measurement, many manufacturers support head circumference, double parietal diameter, abdominal circumference, The automatic measurement of commonly used measurement items such as femoral length has made a great contribution to the improvement of clinical examination efficiency.

However, in ultrasound inspection, image comparison is often involved. In order to facilitate image comparison, multi-window mode is a commonly used image display method. In the multi-window mode, multiple ultrasound images (mainly dual-windows) can be displayed on one screen at the same time, allowing doctors to compare and observe different anatomical structures. Generally, in the multi-window mode, there will be a switch button that allows the user to activate a certain window (hereinafter referred to as the active window), and the image will be scanned in this window in real time, and the remaining windows will display the scanned image.

However, in automatic measurement, the multi-window mode often brings great troubles. When the user starts the automatic measurement of a certain measurement item, the system does not know which window image the user wants to measure. The method is that in the multi-window mode, which window is the current active window, the image of which window is automatically measured. This requires the user to scan an image and immediately perform the measurement, otherwise the automatically measured image is not the image that the doctor wants. However, some doctors are accustomed to lay out all the slices and then perform unified measurements. This requires the user to switch windows to measure, but increases the operation steps.

Summary of the invention

This application is made in order to solve at least one of the above-mentioned problems. The present application provides an automatic measurement method and ultrasound imaging system for anatomical structures, which can obtain an ultrasound image with the measurement item to be measured according to the measurement item to be measured, and perform automatic measurement on the measurement item to be measured in the ultrasound image, Improve the efficiency of automatic measurement of anatomical structures.

In the first aspect, an embodiment of the present application provides an automatic measurement method of an anatomical structure, including:

In an image acquisition step, an ultrasound image is acquired, the ultrasound image is related to at least one anatomical structure of the biological tissue, and the at least one anatomical structure has at least one measurement item;

The measurement item obtaining step is to obtain the measurement item to be measured of the anatomical structure;

An image recognition step of identifying, from the ultrasound image, a recognition image containing a measurement target corresponding to the measurement item to be measured, and the recognition image is at least one of the ultrasound images;

A positioning step, positioning the measurement target in the recognition image;

In an automatic measurement step, the measurement target is measured.

In the second aspect, an embodiment of the present application also provides an automatic measurement method of anatomical structure, including:

An image recognition step of identifying, from the ultrasound image, a recognition image containing a measurement target corresponding to the measurement item to be measured, wherein the recognition image is at least one of the ultrasound images;

In an automatic measurement step, the measurement target in the recognition image is measured.

In a third aspect, an embodiment of the present application also provides an automatic measurement method of anatomical structure, including:

An image acquisition step of acquiring at least two ultrasound images, at least one of the at least two ultrasound images being related to at least one anatomical structure of the biological tissue;

In an image recognition step, a recognition image containing a measurement target corresponding to a measurement item is recognized from the ultrasound image, wherein the measurement item is a measurement item to be measured, and the recognition image is at least one image in the ultrasound image ；

The positioning step is to locate the measurement target in the recognition image.

In an automatic measurement step, the measurement target is measured.

In a fourth aspect, an embodiment of the present application also provides an automatic measurement method of anatomical structure, including:

An embodiment of the present application also provides an ultrasound imaging system, including:

Ultrasound probe, used to transmit ultrasonic waves to biological tissues and receive ultrasonic echoes to obtain ultrasonic echo signals;

A processor, configured to process the ultrasonic echo signal to obtain an ultrasonic image of the biological tissue;

A display for displaying the ultrasound image;

The memory is used to store executable program instructions;

And a processor, configured to execute the executable program instructions, so that the processor executes the automatic measurement method described in any one of the first aspect to the fourth aspect.

The embodiments of the present application provide an automatic measurement method of anatomical structure and an ultrasound imaging system. According to the measurement item to be measured, the ultrasound image containing the measurement item to be measured is identified, and the measurement item to be measured in the ultrasound image is identified. Automatic measurement improves the efficiency of measuring anatomical structures.

Description of the drawings

Fig. 1 shows a schematic block diagram of an ultrasound imaging system according to an embodiment of the present application;

FIG. 2 shows a schematic flowchart of an automatic measurement method of an anatomical structure according to an embodiment of the present application;

Fig. 3 shows a schematic flowchart of an automatic measurement step in an automatic measurement method of an anatomical structure according to an embodiment of the present application;

Fig. 4 shows a schematic flowchart of an automatic measurement method of anatomical structure according to an embodiment of the present application;

Fig. 5 shows a schematic flowchart of an automatic measurement method of anatomical structure according to an embodiment of the present application;

Fig. 6 shows a schematic flowchart of an automatic measurement method of an anatomical structure according to an embodiment of the present application.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the present application more obvious, the exemplary embodiments according to the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited by the exemplary embodiments described herein. Based on the embodiments of this application described in this application, all other embodiments obtained by those skilled in the art without creative work should fall within the protection scope of this application.

In the following description, a lot of specific details are given in order to provide a more thorough understanding of this application. However, it is obvious to those skilled in the art that this application can be implemented without one or more of these details. In other examples, in order to avoid confusion with this application, some technical features known in the art are not described.

It should be understood that this application can be implemented in different forms and should not be construed as being limited to the embodiments presented here. On the contrary, the provision of these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

The purpose of the terms used here is only to describe specific embodiments and not as a limitation of the present application. When used herein, the singular forms "a", "an" and "the/the" are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "composition" and/or "including", when used in this specification, determine the existence of the described features, integers, steps, operations, elements and/or components, but do not exclude one or more other The existence or addition of features, integers, steps, operations, elements, parts, and/or groups. As used herein, the term "and/or" includes any and all combinations of related listed items.

In order to thoroughly understand this application, detailed steps and detailed structures will be presented in the following description to explain the technical solutions proposed by this application. However, in addition to these detailed descriptions, this application may also have other implementations.

Fig. 1 shows a schematic block diagram of an ultrasound imaging system according to an embodiment of the present application. As shown in FIG. 1, the ultrasound imaging system 100 provided in this embodiment includes an ultrasound probe 101, a processor 102, a memory 103 and a display 104. The ultrasonic probe 101 is used to transmit ultrasonic waves to biological tissues and receive ultrasonic echoes to obtain ultrasonic echo signals. The processor 102 is configured to process the ultrasound echo signals to obtain an ultrasound image of the anatomical structure of the biological tissue, and automatically measure the anatomical structure of the biological tissue based on the ultrasound image.

The memory 103 stores executable computer program instructions. When the processor 102 executes the executable computer program instructions, the processor 102 performs automatic measurement of the anatomical structure to obtain a measurement result, for example, a measurement result of a measurement target. The display 104 is used to display the ultrasound image, the measurement result measured by the processor, the measurement target, the measurement item to be measured, and the like.

Referring to FIG. 2, the method for automatically measuring anatomical structures performed by the processor of this embodiment according to an embodiment of the present application provides an exemplary introduction to the method for automatically measuring anatomical structures based on the user designating the measurement item to be measured. . In this embodiment, according to the measurement item specified by the user, the acquired ultrasound image of the anatomical structure automatically recognizes the recognition image with the measurement target corresponding to the measurement item to be measured, and then directly measures the measurement target in the recognition image. . In the whole process, the user only needs to specify the measurement item to be measured, and the user does not need to identify the ultrasonic image containing the measurement target corresponding to the measurement item to be measured according to the measurement item to be measured, and there is no need to manually perform the measurement target in the ultrasonic image. The measurement simplifies the operation process and improves the efficiency of measuring anatomical structures.

Referring to FIG. 2, a schematic flowchart of an automatic measurement method of anatomical structure according to an embodiment of the present application is shown.

The automatic measurement method of anatomical structure provided in this embodiment is used to automatically measure the anatomical structure of biological tissue after processing the ultrasonic echo, as shown in FIG. 2. The method includes:

Step S11: an image acquisition step, acquiring an ultrasound image, the ultrasound image being related to at least one anatomical structure of the biological tissue, and the at least one anatomical structure has at least one measurement item. When performing ultrasonic testing of biological tissues, it is often necessary to observe the anatomical structure of biological tissues. In the case of measuring anatomical structures, the anatomical structure has at least one measurement item. For example, during obstetric ultrasound testing, pregnant women are observed The anatomical structure of the fetus in the body often corresponds to the measurement items such as double parietal diameter, head circumference, abdominal circumference, and femoral length that need to be measured; during abdominal ultrasound examination, the anatomical structure of the liver and kidney of the subject’s abdomen is observed. They correspond to the measurement items of liver size and kidney size respectively. In other embodiments, the anatomical structure of the corresponding biological tissue has different measurement items.

In one embodiment, in the image acquisition step, the processor 102 processes the ultrasonic echo acquired by the ultrasonic probe 101 to generate an ultrasonic image. In one embodiment, the processor 102 processes the ultrasonic echo acquired by the ultrasonic probe 101, generates an ultrasonic image to be stored in the memory 103, and obtains information from the ultrasonic image that has been stored in the memory 103 during the image acquisition step. An ultrasound image of the anatomical structure of the biological body, and the ultrasound image and the processing result are displayed on the display 104. In one embodiment, in the image acquisition step, at least one ultrasound image about the anatomical structure is acquired. In one embodiment, the display 104 displays one or more ultrasound images about the anatomical structure of the biological tissue. For example, two ultrasound images about the head of the fetus are displayed on the display 104 at the same time. In one embodiment, ultrasound images related to different anatomical structures of the biological tissue are displayed on the display 104, for example, ultrasound images related to the head of the fetus and the abdomen of the fetus are simultaneously displayed on the display 104. In one embodiment, the display 104 has multiple display windows, and each display window displays one or more ultrasound images about the anatomy. For example, there are two display windows on the display 104, one of which displays information about the fetus. Ultrasound image of the head, one of the display windows shows the ultrasound image of the fetus' abdomen. In one embodiment, while the ultrasound image is displayed on the display 104, measurement items and measurement results related to the anatomical structure may also be displayed.

Step S12: a measurement item obtaining step, obtaining a measurement item to be tested. In the measurement item acquisition step (S12), the measurement items that need to be measured regarding the anatomical structure of the biological tissue are acquired.

In one embodiment, the measurement items of each of at least one anatomical structure of the biological tissue are displayed on the display 104. In an embodiment, the processor 102 performs the measurement item acquisition step by receiving the measurement item to be measured input by the user. In one embodiment, the processor 102 is connected to an input device, and the user selects the measurement item to be measured from the display 104 through an instruction input by the input device. For example, when performing obstetric ultrasound testing, the display 104 displays the measurement items of the fetal head, fetal abdomen, and placenta during the process of scanning the pregnant woman's abdomen, including: biparietal diameter, occipital frontal diameter, and head circumference , Abdominal circumference, femur length, humerus length, placenta thickness, abdominal transverse diameter, abdominal thickness diameter, neck folds and other measurement items. The user needs to measure the length of the femur of the fetus, and input the instruction to measure the length of the femur through the input device communicatively connected with the processor 102. The processor 102 receives the instruction that the user needs to measure the length of the femur, thereby obtaining the measurement item to be measured. The measurement item is the length of the femur.

In one embodiment, the anatomical structure has a measurement item, which is acquired as the measurement item to be measured in the measurement item acquisition step (S12). In one embodiment, the anatomical structure has two or more measurement items, and at least one of the two or more measurement items can be separately obtained as the measurement to be measured in the measurement item obtaining step (S12). Item, two or more of the two or more measurement items may also be used as the measurement items to be measured at the same time. For example, during the obstetric ultrasound examination, the ultrasound image about the fetus acquired in the image acquisition step, the fetus has measurement items such as double parietal diameter, head circumference, abdominal circumference, and femur length. In the measurement item acquisition step (S12), the user The dual parietal diameter is measured by inputting the input device in communication with the processor 102 to obtain the dual parietal diameter to be measured, or the user can simultaneously measure the dual parietal diameter, head circumference, and femoral length through the input in communication with the processor 102. One instruction, so as to obtain the three measurement items of double parietal diameter, head circumference, and femur length at the same time.

Step S13: an image recognition step, recognizing a recognition image containing a measurement target corresponding to the measurement item to be measured from the ultrasound image, and the recognition image is at least one of the ultrasound images. In the image recognition step (S13), the ultrasound image about the anatomical structure acquired in the image acquisition step is identified, whether it contains the measurement target corresponding to the measurement item specified by the user, and the measurement item specified by the user is included The corresponding measurement target is the recognition image, and the ultrasonic image that does not contain the measurement target corresponding to the measurement item designated by the user is discarded, and the following automatic measurement step is not entered.

In ultrasonic testing, because of the ultrasonic testing of biological tissues, the user is often required to collect the ultrasonic echo of the biological tissues through the probe, and then select the anatomical structure after the ultrasonic image is processed by the ultrasonic system to obtain multiple anatomical structures. The measurement item of is used as the measurement item to be measured for measurement, wherein the ultrasound image may be a cross-sectional image or a three-dimensional image. For example, in the process of two-dimensional ultrasound imaging and measurement of biological tissues, the doctor first lays out the cut surfaces of all tissues and then performs unified measurement, or displays the cut surfaces of the tissues in different windows during the inspection process. Take automatic measurements. In the traditional measurement method, the user is required to manually identify the ultrasound image containing the measurement target corresponding to the measurement item to be measured and then perform the measurement. For example, manually switch the window or select the active window to perform the measurement. This process increases user operations. According to this embodiment, the image recognition step (S13) automatically recognizes the recognition image containing the measurement target corresponding to the measurement item to be measured specified by the user, without requiring the user to manually recognize the ultrasonic image containing the measurement target corresponding to the measurement item to be measured , Automatically measure the automatically recognized image, reduce user operations and improve measurement efficiency.

For example, when performing obstetric ultrasound testing, in the image acquisition step, the ultrasound image about the fetus is acquired. The fetus has three measurement items to be measured: double parietal diameter, head circumference, and femoral length. Among them, the double parietal diameter corresponds to the fetal head The measurement target of the parietal bones on both sides of the fetus and the head circumference correspond to the measurement target of the occipital bone of the fetal head to the forehead nasal root, and the length of the femur corresponds to the measurement target of the fetal femur. In the measurement item acquisition step, the acquisition user needs to measure the fetal femur length, and the ultrasound image about the fetus contains the ultrasound image including the head and the ultrasound image including the femur, which is required in the image recognition step The ultrasonic image including the head and the ultrasonic image including the femur are recognized to identify the ultrasonic image including the femur, and the subsequent automatic measurement step is performed on the recognized ultrasonic image including the femur.

In one embodiment, the recognition image is a part of an ultrasound image. For example, it may be a partial area of a certain ultrasound image. When performing obstetric ultrasound testing, in the image acquisition step, the ultrasound image about the fetal head is acquired. The fetal head has two measurement items to be measured: double parietal diameter and head circumference. Among them, the double parietal diameter corresponds to the fetal head The measurement target of the parietal bones on both sides and the head circumference correspond to the measurement target of the occipital bone of the fetus head to the root of the forehead and nose. The measurement item of double parietal diameter is obtained in the measurement item acquisition step, and the ultrasonic image recognized in the image recognition step Only the area of the parietal bones on both sides of the fetus’s head is included. In one embodiment, one or more measurement items to be measured are obtained in the measurement item acquisition step (S12), and multiple recognition images are recognized in the image recognition step (S13), wherein at least two recognition images contain the same measurement item to be tested. For the measurement target corresponding to the measurement item, in the subsequent automatic measurement step, the measurement target corresponding to the measurement item to be measured in each recognition image is measured, and the measurement target in each recognition image is measured. The results are averaged, and the average is the measurement result for the measurement item to be tested. For example, when performing obstetric ultrasound testing, in the image acquisition step, multiple ultrasound images about the fetus are acquired. In the measurement item acquisition step, the user needs to measure the fetal femur length. In the recognition step, two ultrasound images including the femur are identified from multiple ultrasound images, and both of the ultrasound images including the femur are recognized images, and then the recognized ultrasound images including the femur are automatically measured. In the step, the two identification images are measured, and the measurement results of the two identification images are averaged, and the average value is the measurement result for the measurement item to be measured. Of course, in other implementation manners, in addition to calculating the average value, the final measurement result may also be determined by weighting or calculating the variance, which is not specifically limited here. In one embodiment, at least two ultrasound images related to the anatomical structure of the biological tissue are acquired in the image acquisition step (S11); the measurement item acquisition step (S12) acquires one or more measurement items of the aforementioned anatomical structure; the image recognition step In (S13), the ultrasound images are classified according to the measurement items of each of the at least one anatomical structure of the biological tissue, so as to obtain the ultrasound image containing the measurement target corresponding to the measurement item; Select the ultrasound image with the measurement target corresponding to the measurement item to be measured from the ultrasound images.

In one embodiment, the foregoing step of classifying the ultrasound image includes: comparing the image feature of the ultrasound image with the image feature of the database image in a preset database, wherein the database image contains at least one of the The measurement target corresponding to the measurement item of at least one of the anatomical structures of the biological tissue, when the image feature of the ultrasound image matches the image feature of the database image, the ultrasound image contains the The measurement target contained in the database image.

For example, in obstetric ultrasound inspection, multiple ultrasound images are acquired in the image acquisition step (S11), including ultrasound images of the fetus head and ultrasound images of the fetus abdomen. The measurement items of the fetal head include: double parietal diameter, head circumference, etc., and the measurement items of the fetal abdomen include: abdominal circumference, abdominal transverse diameter, and abdominal thickness diameter. The measurement item to be measured obtained in the measurement item acquisition step (S12) is the head circumference. In the image recognition step (S13), the image acquisition step (S11) is based on the measurement items on the head of the fetus and the measurement items on the abdomen of the fetus. ) The acquired multiple ultrasound images are classified, and the measurement target containing the parietal regions on both sides of the fetal head corresponding to the double parietal diameter and the measurement target of the head circumference corresponding to the occipital bone of the fetal head to the forehead nasal root region are obtained. And an ultrasound image containing the measurement target of the abdomen corresponding to the abdominal circumference, the transverse diameter of the abdomen, and the thickness of the abdomen Specifically, each of the multiple ultrasound images is compared with the image features of the database image in the preset database. The database image contains at least the measurement targets (such as the parietal bones on both sides of the head, the occipital bones to the forehead, nose, and abdomen). In one item, when the image feature of the ultrasound image matches the image feature of the database image, the ultrasound image corresponds to the measurement target contained in the database image. In the foregoing steps, after the multiple ultrasound images acquired in the image acquisition step (S11) are classified, the measurement item to be measured acquired in the measurement item acquisition step (S12) is directly used as the head circumference, and from the classified ultrasound images Obtaining an ultrasound image from the occipital bone of the head to the root of the forehead and nose is the recognition image. In one embodiment, the multiple ultrasound images acquired in the image acquisition step (S11) are classified in the image recognition step (S13) on the display 104 and the ultrasound images containing measurement targets corresponding to different measurement items are displayed.

In one embodiment, the recognition image recognized in the image recognition step (S13) is displayed on the display 104. In one embodiment, the processor 102 is connected to the input device, the user selects the measurement item to be measured from the display 104 through the instruction input by the input device, and the processor 102 obtains the image acquisition step (S11) according to the measurement item to be measured selected by the user. After performing the image recognition step (S13) for multiple ultrasound images, the recognized recognition images are displayed on the display 104. In one embodiment, multiple ultrasound images are displayed on the display 104, and at the same time, the recognition images are displayed in a manner that distinguishes other ultrasound images, for example, displayed in a highlighted manner. For example, during the obstetric ultrasound examination, the two ultrasound images of the fetus in the image acquisition step (S11) are displayed on the display 104, one of which contains the fetal head and the occipital bone to the forehead and nasal root region, and the other contains the fetus. For the abdomen, according to the user's instruction to measure the head circumference of the fetus, the processor 102 performs the image recognition step (S13) according to the measurement head circumference selected by the user, and then converts the identified head occipital bone to the forehead nasal root. The recognition image of the measurement target of the area is displayed on the display 104 in such a way that it is distinguished from the ultrasound image including the abdomen.

In one embodiment, the image recognition step (S13) includes: comparing the image feature of the ultrasound image with the image feature of the database image containing the measurement item to be measured in the preset database, and judging the image feature of the ultrasound image and the image of the database image Whether the feature matches, when the image feature of the ultrasound image matches the image feature of the database image, it is determined that this ultrasound image is a recognition image containing the measurement target corresponding to the measurement item to be measured. The database image included in the preset database is an image calibrated for the measurement items of the anatomical structure, and it contains measurement targets corresponding to the measurement items of the anatomical structure.

In one embodiment, at least two ultrasound images related to the anatomical structure of the biological tissue are acquired in the image acquisition step (S11); the measurement item acquisition step (S12) acquires one measurement item of the aforementioned anatomical structure; the image recognition step (S13) Compare each ultrasound image acquired in the image acquisition step with the database image of the measurement target corresponding to the aforementioned measurement item in the preset database, and determine whether the image feature of the currently compared ultrasound image matches the image feature of the database image, If it matches, it is determined that the currently compared ultrasound image is an identification image; if it does not match, the ultrasound image is discarded, thereby determining the identification image containing the measurement target corresponding to the measurement item to be measured from the multiple ultrasound images.

For example, during abdominal ultrasound detection, in the image acquisition step (S11), two ultrasound images of human organs are acquired. Human organs include liver, kidney, etc., where the liver has a measurement item for liver size, and the kidney has a measurement for kidney size. Item, the measurement item to be measured acquired in the measurement item acquisition step (S12) is the size of the liver, and in the image recognition step (S13), each of the two ultrasound images of the human organs acquired in the image acquisition step (S11) Compare with the database image containing the liver in the preset database (also the ultrasound image about the liver), determine whether the image feature of the ultrasound image matches the image feature of the database image, if it matches, it will be determined as the recognition image, if it does not match , It will be discarded. In one embodiment, in the image acquisition step (S11), two or more ultrasound images related to the anatomical structure of the biological tissue are acquired, the biological tissue has one or more anatomical structures, and one of the one or more anatomical structures There are two or more measurement items; the measurement item to be measured obtained in the measurement item acquisition step (S12) is two or more of the two or more measurement items of the aforementioned anatomical structure; the image recognition step (S13) also Including: classifying the recognition image to obtain the recognition image containing the measurement target corresponding to each measurement item to be measured. Since the anatomical structure of biological tissue often has multiple measurement items, and multiple measurement items need to be measured at the same time, in the image recognition step (S13), the process of matching the image characteristics of the ultrasound image with the image characteristics of the database image It is also necessary to classify the recognition images according to the measurement items to be tested to obtain the recognition image of the measurement target corresponding to each measurement item to be tested.

For example, in the obstetric ultrasound examination, two or more ultrasound images about the fetus are acquired in the image acquisition step (S11). The fetus has measurement items such as double parietal diameter, head circumference, abdominal circumference, and femoral length. The measurement items to be measured acquired in (S12) are head circumference and abdominal circumference. In the image recognition step (S13), each of the ultrasound images of two or more fetuses of the fetus acquired in the image acquisition step (S11) is separately Compare with the database image of the head and abdomen (also about the ultrasound image of the fetus) in the preset database, where the comparison process is also based on the image features of the database image containing the head and the image feature of the database image containing the abdomen The recognition images are classified to determine the head recognition image corresponding to the measurement item of head circumference and the abdomen recognition image corresponding to the measurement item of abdominal circumference.

In one embodiment, a machine learning algorithm is used to learn the image features of the database images in the preset database that can distinguish different measurement items. At the same time, the machine learning method is used to extract the image features of the ultrasound image acquired in the image acquisition step (S11), and The learned image feature of the database image is matched with the image feature of the ultrasound image, and the ultrasound image matching the learned image feature is obtained as the recognition image. In the case of multiple measurement items to be measured, the ultrasound images are classified according to the learned image characteristics that can distinguish different measurement items, so as to classify the ultrasound images according to the measurement items of the anatomical structure, so that the ultrasound images are classified with each other. The corresponding recognition image of the measurement item to be tested is recognized. Among them, the method of extracting features by machine learning algorithms includes, but is not limited to, principal component analysis (Principle Components Analysis, PCA) method, linear discriminant analysis (Linear Discriminant Analysis, LDA) method, Harr feature extraction method, texture feature extraction method, etc. Match the image features of the ultrasound images extracted by the machine learning algorithm with the image features in the preset database to classify the ultrasound images. The classification discriminators used include but are not limited to nearest neighbor (KNN), Support vector machine (Suport vector maehine, SVM) random forest, neural network and other discriminators.

In one embodiment, a deep learning method is used to construct a stacked convolutional layer and a fully connected layer to learn image features of database images in a preset database, and learn image features that can distinguish different measurement items. According to these images Feature recognition is the recognition image in the ultrasound image. In the case of having at least two measurement items to be measured, the ultrasound images are classified according to the learned image characteristics, the images with the aforementioned measurement items that can be distinguished in the ultrasound images are recognized, and the recognized images are the recognition images. Deep learning methods include, but are not limited to, VGG network, ResNet residual network, Inception module, AlexNe ot deep network, etc.

Step S14: an automatic measurement step, measuring the measurement target in the recognition image. In the automatic measurement step (S14), the measurement target in the recognition image is measured to obtain the measurement result of the anatomical structure. In ultrasonic testing, when measuring different measurement items, the measurement methods are different. For example, in obstetric ultrasound testing, the measurement of head circumference usually uses an ellipse to wrap the fetal skull halo, and the abdominal circumference uses an ellipse to wrap Live the abdomen of the fetus, and measure the distance between the two ends of the femur with a line segment for the femur length. In the automatic measurement step (S14), the target fitting method is used for automatic measurement.

Referring to FIG. 3, a flowchart of an automatic measurement step according to an embodiment of the present application is shown. As shown in Figure 3, the automatic measurement step (S14) includes:

Step S141: Extract the contour of the measurement target corresponding to the measurement item to be measured by using an edge detection algorithm. Edge detection algorithms include, but are not limited to: using Sobel operator, canny operator, etc., to detect the contour of the measurement item to be measured based on the pixel points and gray-scale weighted values of the ultrasound image.

Step S142: Fit the contour of the measurement target corresponding to the measurement item to be measured to obtain a fitting equation corresponding to the measurement item to be measured. Detecting algorithms such as straight lines, circles, ellipses, etc. are used to fit the contours of the measurement items to be measured to obtain the fitting equations. Fitting algorithms include but are not limited to least squares estimation, Hough transform, Randon transform, Ransac and other algorithms.

Step S143: Determine the measurement result through the fitting equation. The measurement result is determined according to the fitting equation obtained by the fitting algorithm in the above steps. If the fitting equation circle or ellipse equation is obtained in the above steps, it is the result of automatic measurement. If the fitting equation obtained in the above steps is a straight line, the end point can be further located by combining the gray change of the end point of the measurement item to be measured to realize automatic measurement. Taking the measurement of femur length in obstetric ultrasound testing as an example, the femur appears as a bright linear structure. After detecting the straight line where the femur is located, two points with the largest gray gradient of the femur can be detected on the straight line as the two end points of the femur.

In one embodiment, the measurement result measured in the automatic measurement step (S14) is displayed on the display 104. In one embodiment, the processor 102 performs an automatic measurement step (S14) on the recognition image recognized in the image recognition step (S13), and displays the measurement result on the recognition image displayed on the display 104. For example, when performing an obstetric ultrasound inspection, the processor 102 performs an image recognition step (S13) according to the measurement head circumference selected by the user, and then recognizes the head occipital bone to the forehead nasal root region corresponding to the measurement item including the head circumference. The recognition image of this measurement target is displayed on the display 104 in a way that highlights the ultrasonic image of the measurement target that is different from the parietal region on both sides of the fetal head corresponding to the measurement item of double parietal diameter, and the subsequent automatic measurement step ( The specific numerical value of the head circumference obtained in S14) is displayed in the upper right corner of the recognition image with the measurement item of head circumference recognized in the image recognition step (S13).

Referring to FIG. 4, the method for automatically measuring anatomical structures by a processor according to an embodiment of the present application provides an exemplary introduction to the method for automatically measuring anatomical structures based on the user designating a measurement item to be measured. In this embodiment, a method for automatically measuring an anatomical structure based on the user designating a measurement item to be measured is provided. According to the measurement item specified by the user, the acquired ultrasound image of the anatomical structure automatically recognizes the recognition image containing the measurement target corresponding to the measurement item to be measured, and then directly measures the measurement target in the recognition image. In the whole process, the user only needs to specify the measurement item to be measured, and the user does not need to identify the ultrasonic image containing the measurement target corresponding to the measurement item to be measured according to the measurement item to be measured, and there is no need to manually perform the measurement target in the ultrasonic image. The measurement simplifies the operation process and improves the efficiency of measuring anatomical structures. The difference from the embodiment shown in FIG. 2 is that in this embodiment, a positioning step is added after the image recognition step to eliminate the influence of the surrounding structure of the measurement target on the measurement result in the measurement step. Next, referring to FIG. 4, the automatic measurement method of the anatomical structure of this embodiment will be exemplified.

Referring to FIG. 4, a schematic flowchart of an automatic measurement method of anatomical structure according to an embodiment of the present application is shown, in which an image recognition step (S21), a measurement item acquisition step (S22), and an image recognition step (S23) are shown. It is consistent with the image recognition step (S11), the measurement item acquisition step (S12) and the image recognition step (S13) shown in FIG. 2, except that the positioning step (S24) is added after the image recognition step (S23), The measurement target is measured in the automatic measurement step (S25). The measurement target in the positioning step (S24) and the automatic measurement step shown in FIG. 4 will be described in detail below.

Step S24: a positioning step, positioning the measurement target in the recognition image. In the above image recognition step (S13), only the recognition image containing the measurement target corresponding to the measurement item to be measured is obtained, and the position corresponding to the measurement target in the actual measurement is not known, and the measurement target in the recognition image is directly determined. The measurement requires the detection of the entire image, and the obtained edge detection result is easily affected by the structure around the measurement target. For this reason, in the positioning step (S24), the measurement target is positioned, and then the measurement target is fitted to the measurement target in the automatic measurement step, which can reduce the influence of the surrounding structure of the measurement item and make the measurement result more accurate.

In one embodiment, in the positioning step (S24), the image feature of the recognition image is compared and analyzed with the image feature of the database image containing the measurement target corresponding to the measurement item to be measured in the preset database, so as to The measurement target is obtained by locating the measurement target in the recognition image, wherein the database image contains a calibration result corresponding to the measurement target, and the measurement target is an area consistent with the calibration result.

In an embodiment, the calibration result includes the ROI box of the measurement target corresponding to the measurement target.

In an embodiment, the calibration result includes the ROI box of the measurement target corresponding to the measurement item to be measured. The positioning step includes: extracting the image features in the sliding window by using a method based on the sliding window, comparing the image features in the sliding window with the image features of the calibration result, and judging the difference between the image features in the sliding window and Whether the image feature of the calibration result matches, and when the image feature in the sliding window matches the image feature of the calibration result, it is determined that the current sliding window is the measurement target.

In one embodiment, in the positioning step (S24), a machine learning algorithm is used to learn the image features in the ROI box of the calibration result of the database image in the preset database, where the learned image features in the ROI box are the measured The image feature of the calibration result that distinguishes the ROI area and non-ROI area of the target. At the same time, a machine learning algorithm is used to extract the image features in the sliding window obtained when the sliding window traversal is performed on the recognition image recognized in the image recognition step (S13). Among them, the method of machine learning algorithm to extract features includes but not limited to principal component analysis (Principle Components Analysis, PCA) method, linear discriminant analysis (Linear Discriminant Analysis, LDA) method, Harr feature extraction method, texture feature extraction method, etc.

In one embodiment, the calibration result includes the ROI frame of the measurement target corresponding to the measurement item to be measured, and the positioning step includes: according to the calibration result in the database image containing the measurement target The frame regression processing is performed on the recognized image to obtain a frame area, and the frame area is the measurement target.

In one embodiment, a deep learning method is used to construct a stacked base convolutional layer and a fully connected layer in the ROI box of the calibration result of the database image containing the measurement target corresponding to the measurement item in the preset database. Image feature learning and parameter regression, the learned image feature in the ROI frame is the image feature of the calibration result that distinguishes the ROI area and the non-ROI area of the measurement target. According to the learned image characteristics, the neural network algorithm directly returns the border area of interest in the recognized image, and this border area is the measurement target to be measured. Among them, neural network algorithms include but are not limited to R-CNN, Fast R-CNN, Faster-RCNN, SSD, YOLO and other detection algorithms.

In one embodiment, the calibration result includes a mask for accurately segmenting the measurement target, and the positioning step includes: according to image characteristics of the calibration result in the database image containing the measurement target , Using a semantic segmentation algorithm to identify the segmentation mask of the measurement target that is consistent with the calibration result in the recognition image.

In one embodiment, a deep learning method is used to perform end-to-end semantic segmentation to segment the recognition image. Specifically, construct a stacked base convolutional layer or use a deconvolutional layer to sample a mask that accurately divides the measurement target corresponding to the measurement item in the preset database, and directly obtain the AND contained in the recognition image according to the sampling result. The segmentation mask of the measurement target corresponding to the measurement item to be tested.

After the measurement target in the recognition image is obtained in the positioning step (S24), in the automatic measurement step, the measurement target is measured (S25) to realize the automatic measurement process of the anatomical structure. In the automatic measurement step, measuring a measurement target (S25) includes performing target fitting on the measurement target to obtain a fitting equation of the measurement target; and determining the measurement result of the measurement target through the fitting equation.

In one embodiment, the calibration result in the positioning step (S24) includes the ROI frame of the measurement target corresponding to the measurement item to be measured. During the process of measuring the measurement target (S25), the frame area of the measurement target is located in the recognition image. Target fitting is performed on the measurement target inside, and the fitted line, circle or ellipse equations are obtained, and the measurement results are obtained by calculating the aforementioned equations. Performing target fitting on the measurement target in the frame area of the measurement target can reduce the interference of other structures outside the frame area on the target fitting, and improve the accuracy of the measurement result.

In one embodiment, the calibration result in the positioning step (S24) includes a mask for accurately segmenting the measurement target corresponding to the measurement item, and the recognition image is compared with the calibration in the process of measuring the measurement target (S25). The edge of the segmented mask of the measurement target with consistent results is fitted to the target, and the equations such as a straight line, a circle or an ellipse are fitted, and the measurement result is obtained by calculating the foregoing equations. Target fitting is performed on the edge of the segmentation mask of the measurement target that is consistent with the calibration result in the recognition image, which can reduce the fitting error of the target fitting and improve the accuracy of the measurement result.

This embodiment provides a method for automatically measuring an anatomical structure based on the user designating a measurement item to be measured. According to the measurement item specified by the user, the acquired ultrasound image of the anatomical structure automatically recognizes the identification image containing the measurement target corresponding to the measurement item to be measured, and then directly measures the measurement item in the identification image. In the whole process, the user only needs to specify the measurement item to be measured, and the user does not need to identify the ultrasonic image containing the measurement target corresponding to the measurement item to be measured according to the measurement item to be measured, and there is no need to manually perform the measurement target in the ultrasonic image. The measurement simplifies the operation process and improves the efficiency of measuring anatomical structures.

Referring to FIG. 5, an automatic measurement method of an anatomical structure according to an embodiment of the present application provides an exemplary introduction to a method for automatic measurement of an anatomical structure based on a user designating a measurement item to be measured. This embodiment provides a method for measuring all the measurability of the anatomical structure based on the user not specifying the measurement item to be measured. The user does not need to specify. After the ultrasound image of the anatomical structure is obtained, the ultrasound image is directly automatically recognized to identify the recognition image containing the measurement target, and the measurement target in the recognition image is automatically measured, where the measurement target corresponds to The measurement items of the anatomical structure. In the whole process, the user only does not need to perform any operation on the measurement items, which further simplifies the operation process and improves the efficiency of measuring the anatomical structure.

Referring to FIG. 5, there is shown a schematic flowchart of an automatic measurement method of an anatomical structure according to an embodiment of the present application.

The automatic measurement method of anatomical structure provided in this embodiment is used to automatically measure the anatomical structure of the biological tissue to be tested after processing the ultrasonic echo, as shown in FIG. 5. The method includes:

Step S31: an image acquisition step, acquiring at least two ultrasound images, at least one of the ultrasound images being related to at least one anatomical structure of the biological tissue. The automatic measurement method of anatomical structure provided in this embodiment is used to identify an ultrasonic image containing a measurement target from at least two ultrasonic images related to the anatomical structure of a biological tissue, and measure the measurement target in the ultrasonic image, wherein the measurement The target corresponds to the measurement items possessed by the anatomical structure.

In ultrasonic testing, when performing ultrasonic testing on biological tissues, it is often necessary to collect ultrasonic echoes of biological tissues through a probe, and after processing to obtain multiple ultrasound images about the anatomical structure of biological tissues, select The measurement items of the anatomical structure are measured as the measurement items to be measured. In traditional measurement methods, the measurement items to be measured are often specified by the user, and the user manually recognizes the ultrasound image containing the measurement target corresponding to the measurement item to be measured and then performs the measurement. This process increases the need for the user to specify the measurement item operation; because the measurement item that the user wants to measure is often fixed, it is necessary to measure all the measurement items of the anatomical structure. In this embodiment, the anatomy is not specified based on the user. All measurement items of the structure are measured, which reduces the user's operation steps and simplifies the measurement operation of the anatomical structure.

Step S32: an image recognition step, recognizing a recognition image containing a measurement target corresponding to a measurement item from the ultrasound image, wherein the recognition image is at least one of the ultrasound images. Since this embodiment measures all the measurement items that the anatomical structure has, the measurement targets in the identification image of the measurement target corresponding to the measurement item that contains the anatomical structure identified from the ultrasound image need to be measured, that is, the measurement of the anatomical structure All items are to-be-tested measurement items.

In an embodiment, the at least one anatomical structure of the biological tissue has at least one feature measurement item, and the image recognition step (S32) includes: combining the image feature of the ultrasound image with a preset database containing the at least one feature. The image feature of the database image of the measurement target corresponding to any one of the feature measurement items is compared to determine whether the image feature of the ultrasound image matches the image feature of the database image, and when the image feature of the ultrasound image matches the image feature of the database When the image features of the image are matched, it is determined that the ultrasound image is the recognition image, the measurement item corresponding to the measurement target contained in the recognition image is the measurement item to be measured, and the measurement target corresponds to the recognition image The image feature matches the measurement target contained in the database image.

For example, during abdominal ultrasound testing, two ultrasound images are acquired in the image acquisition step (S31), one of which is an ultrasound image about the liver, and the other is an ultrasound image obtained by a doctor's misoperation during the operation , It has nothing to do with the anatomical structure of any biological tissue. The liver has this feature measurement item of liver size. In the image recognition step (S32), the image acquisition step (S31) acquires each of the two ultrasound images and the database image containing the liver size in the preset database (also about the liver). The ultrasound image of the liver) is compared to determine whether the image feature of the liver ultrasound image matches the image feature of the database image. If it matches, it will be determined as a recognition image. If it does not match, it will be discarded; the matching recognition image contains The liver of is the measurement target, and the characteristic measurement item corresponding to the size of the liver is used as the measurement item to be measured.

In an embodiment, any one or two of the at least one anatomical structure of the biological tissue has at least two characteristic measurement items, and the identification image contains at least two measurement targets corresponding to the at least two measurement items to be measured, respectively , The image recognition step (S32) includes: classifying the at least two measurement targets in the recognition image, so that each of the at least two measurement targets respectively corresponds to the at least two feature measurement One of the items. Since the anatomical structure of biological tissues often have multiple measurement items, and multiple measurement items often appear in the same ultrasound image, after the image recognition step (S32) recognizes the recognition image containing multiple measurement targets, it is necessary to The measurement targets contained in the recognition image are classified, that is, the measurement items contained in the recognition image are classified according to the at least two measurement items of the anatomical structure, and one-to-one correspondence with at least two measurement items of the anatomical structure is obtained. Measurement target.

For example, in the obstetric ultrasound examination, two or more ultrasound images about the fetus are acquired in the image acquisition step (S31). The fetus has measurement items such as double parietal diameter, head circumference, abdominal circumference, and femur length. (S31) Each of the acquired ultrasound images of two or more fetuses of the fetus is respectively a database containing measurement targets corresponding to at least one of double parietal diameter, head circumference, abdominal circumference, and femoral length in the preset database The images (also about the ultrasound images of the fetus) are compared to determine the recognition image. Since the fetus has four measurement items of double parietal diameter, head circumference, abdominal circumference, and femoral length, the double parietal diameter and head circumference correspond to the measurement target (the double parietal diameter corresponds to this measurement target of the parietal area on both sides of the fetal head The measurement target from the occipital bone to the forehead nasal root area corresponding to the head circumference often appears in the same ultrasound image, and the identification of the measurement target corresponding to the two measurement items including double parietal diameter and head circumference is determined. After the image is imaged, it is necessary to further distinguish the measurement targets corresponding to the two measurement items to be measured in the recognition image to determine which measurement target corresponds to the head circumference and which measurement target corresponds to the double parietal diameter in the recognition image.

In one embodiment, a machine learning algorithm is used to learn the image features of the database images in the preset database that can distinguish different measurement items. At the same time, the machine learning method is used to extract the image features of the ultrasound image acquired in the image acquisition step (S31), and The learned image feature of the database image is matched with the image feature of the ultrasound image, and the ultrasound image matching the learned image feature is obtained as the recognition image. In the case of multiple measurement items to be measured, the measurement targets are classified according to the learned image features that can distinguish different measurement items, so as to realize the recognition of the measurement targets in the recognition image according to the measurement items possessed by the anatomical structure. Among them, the method of extracting features by machine learning algorithms includes, but is not limited to, principal component analysis (Principle Components Analysis, PCA) method, linear discriminant analysis (Linear Discriminant Analysis, LDA) method, Harr feature extraction method, texture feature extraction method, etc. According to the learned image features that can distinguish different measurement targets, the measurement targets are classified to classify the measurement targets in the recognition image. The classification discriminators used include but are not limited to nearest neighbor classification (K nearest neighbor, KNN), support Vector machine (Suport vector maehine, SVM) random forest, neural network and other discriminators.

In one embodiment, a deep learning method is used to construct a stacked convolutional layer and a fully connected layer to learn image features of database images in a preset database, and learn image features that can distinguish different measurement targets. According to these images Feature recognition is the recognition image in ultrasound images. In the case that the anatomical structure has at least two measurement items to be measured, the measurement targets in the recognition image are classified according to the learned image features. Among them, deep learning methods include, but are not limited to, VGG network, ResNet residual network, Inception module, AlexNe ot deep network, etc.

Step S25: an automatic measurement step, measuring the measurement target in the recognition image. In the automatic measurement step (S25), the measurement target in the recognition image is measured to obtain the measurement result of the anatomical structure. In the case of the recognition image that contains the measurement target recognized in the image recognition step (S23), and the measurement target is classified when the recognition image contains multiple measurement targets, the edge detection is directly performed on the measurement target in the recognition image. The method of performing target fitting to obtain the target fitting equation is used for automatic measurement.

In one embodiment, the automatic measurement step (S25) includes:

The contour corresponding to the measurement target is extracted by an edge detection algorithm. Edge detection algorithms include, but are not limited to: using Sobel operator, canny operator, etc. to detect the contour of the measurement target based on the pixel points and gray-scale weighted values of the ultrasound image.

Fitting the contour corresponding to the measurement target to obtain a fitting equation corresponding to the measurement target. Detecting algorithms such as lines, circles, ellipses, etc. are used to fit the contour of the measurement item to be measured to obtain a fitting equation. The fitting algorithms include but are not limited to least squares estimation, Hough transform, Randon transform, Ransac and other algorithms.

The measurement result of the measurement target is determined by the fitting equation. The measurement result is determined according to the fitting equation obtained by the fitting algorithm in the above steps. If the fitting equation circle or ellipse equation is obtained in the above steps, it is the result of automatic measurement. If the fitting equation obtained in the above steps is a straight line, the end point can be further located in conjunction with the gray change of the end point of the measurement target to realize automatic measurement. Taking the measurement of femur length in obstetric ultrasound testing as an example, the femur appears as a bright linear structure. After detecting the straight line where the femur is located, two points with the largest gray gradient of the femur can be detected on the straight line as the two end points of the femur.

Referring to FIG. 6, the method for automatically measuring anatomical structures performed by the processor of this embodiment according to an embodiment of the present application provides an exemplary introduction to the method for automatically measuring anatomical structures based on the user designating the measurement item to be measured. . This embodiment provides a method for measuring all the measurability of the anatomical structure based on the user not specifying the measurement item to be measured. The user does not need to specify. After the ultrasound image of the anatomical structure is obtained, the ultrasound image is automatically recognized to identify the recognition image containing the measurement target, and the measurement target in the recognition image is automatically measured, where the measurement target corresponds to The measurement items of the anatomical structure. In the whole process, the user does not need to perform any operation on the measurement items, which further simplifies the operation process and improves the efficiency of measuring the anatomical structure. The difference from the embodiment shown in FIG. 5 is that in this embodiment, a positioning step is added after the image recognition step to eliminate the influence of the surrounding structure of the measurement target on the measurement result in the measurement step. Next, referring to FIG. 6, the automatic measurement method of the anatomical structure of this embodiment will be exemplified.

Referring to FIG. 6, there is shown a schematic flowchart of an automatic measurement method of an anatomical structure according to an embodiment of the present application. Among them, the image recognition step (S41) and the image recognition step (S42) are consistent with the image recognition step (S31) and the image recognition step (S32) shown in FIG. 5. The difference is that after the image recognition step (S42) Add a positioning step (S43), and then perform an automatic measurement step (S44). The following describes the measurement item positioning step (S42) and the automatic measurement step (S44) shown in FIG. 4 in detail.

Step S43: a positioning step, positioning the measurement target in the recognition image.

In the image recognition step (S42), only the recognition image containing the measurement target is obtained, and the position of the measurement target in the actual measurement is not known. Directly measuring the measurement items in the recognition image requires detecting the entire image, and the obtained edge detection results are easily affected by the structure around the measurement items. For this reason, in the positioning step (S43), the measurement target in the recognition image is located, and then the measurement target is fitted to the measurement target in the automatic measurement step, which can reduce the influence of the structure around the measurement target and make the measurement result more accurate .

In one embodiment, in the positioning step (S43), the image feature of the recognition image is compared and analyzed with the image feature of the database image corresponding to the measurement item to be measured in the preset database, so as to compare all the features of the image. The measurement target in the recognition image is positioned, wherein the database image contains a calibration result corresponding to the target to be measured, and the measurement target is an area consistent with the calibration result.

In an embodiment, the calibration result includes the ROI box of the measurement target.

In one embodiment, in the positioning step (S43), a machine learning algorithm is used to learn the image features in the ROI frame of the calibration result of the database image in the preset database, wherein the learned image feature in the ROI frame is the measurement The image feature of the calibration result that distinguishes the ROI area and non-ROI area of the target. At the same time, a machine learning algorithm is used to extract the image features in the sliding window obtained when the sliding window traversal is performed on the recognition image recognized in the image recognition step (S13). Among them, the method of extracting features by machine learning algorithms includes, but is not limited to, principal component analysis (Principle Components Analysis, PCA) method, linear discriminant analysis (Linear Discriminant Analysis, LDA) method, Harr feature extraction method, texture feature extraction method, etc.

In one embodiment, the calibration result includes the ROI frame of the measurement target corresponding to the measurement item to be measured, and the positioning step includes: according to the database containing the measurement target corresponding to the measurement item to be measured For the image feature of the calibration result in the image, frame regression processing is performed on the recognition image to obtain a frame area, and the frame area is the measurement target.

In one embodiment, a deep learning method is used to construct a stacked base convolutional layer and a fully connected layer to learn and parameterize the image features in the ROI frame of the calibration result of the database image corresponding to the measurement item to be measured in the preset database In regression, the learned image feature in the ROI frame is the image feature of the calibration result that distinguishes the ROI area of the measurement target from the non-ROI area. According to the learned image characteristics, the neural network algorithm directly returns the border area of interest in the recognized image, and this border area is the measurement target to be measured. Among them, neural network algorithms include but are not limited to R-CNN, Fast R-CNN, Faster-RCNN, SSD, YOLO and other target detection algorithms.

In an embodiment, the calibration result includes a mask for accurately segmenting the measurement target corresponding to the measurement item, and the positioning step includes: according to the measurement target corresponding to the measurement item to be measured For the image features of the calibration result in the database image, a semantic segmentation algorithm is used to identify the segmentation mask of the measurement target in the recognition image that is consistent with the calibration result.

In one embodiment, an end-to-end semantic segmentation network method based on a deep learning method is used to perform network segmentation on the recognition image. Specifically, construct a stacked base convolution layer or use a deconvolution layer to sample the mask for precise network segmentation of the calibration area corresponding to the measurement target in the preset database, and directly obtain the contained image from the recognition image according to the sampling result. The segmentation mask of the measurement target corresponding to the measurement item to be measured. Semantic segmentation networks used include, but are not limited to, Fully Convolutional Networks (FCN), U-Net Convolutional Networks (U-Net Convolutional Networks), etc.

In an embodiment, the recognition image obtained in the image recognition step (S42) includes at least two measurement targets corresponding to the measurement items to be measured, and the positioning step (S43) further includes: Classification is performed so that the measurement target corresponds to the at least two measurement items to be measured in a one-to-one correspondence. Since the recognition image obtained in the image recognition step (S42) includes at least two measurement targets corresponding to the measurement items to be measured, it is not only necessary to locate the position of the measurement target in the recognition image during the process of locating the measurement items. It also distinguishes the measurement target category to which each location belongs, and obtains the measurement item corresponding to the measurement result after the measurement target is automatically measured in the subsequent automatic measurement step.

For example, in obstetric ultrasound detection, the recognition image of the fetus recognized in the image recognition step (S42) also contains the measurement target of the occipital bone of the fetal head corresponding to the head circumference and the fetal head corresponding to the double parietal diameter. The two measurement targets of the parietal area on both sides of the fetus. During the measurement process after positioning, it is not known whether the measurement is head circumference or double parietal diameter. It is necessary to measure the occipital bone of the fetal head to the forehead nasal root area. Distinguish this measurement target from the parietal area on both sides of the fetal head, and obtain measurement targets corresponding to the head circumference and double parietal diameter respectively.

In one embodiment, the image feature of the measurement target is compared with the image feature of the database image that is calibrated for the measurement target corresponding to the measurement item to be measured in a preset database, to Determine whether the image feature of the measurement target matches the image feature of the calibration result, and when the image feature of the measurement target matches the image feature of the calibration result, it is determined that the measurement target corresponds to the to-be-measured Measurement items.

Still taking the two measurement items of head circumference and abdominal circumference in one recognition image during obstetric ultrasound detection as an example for explanation, the image of each measurement target determined in the recognition image of the fetus recognized in the positioning step (S43) The feature is compared with the image feature of the calibration result of the head circumference calibration that contains the database image of the head circumference in the preset database. When the image feature of the measurement target matches the image feature of the calibration result of the head circumference calibration, the current is judged The measurement target for comparison is the head circumference.

In one embodiment, the machine learning algorithm is collected to learn the image features of the database image in the preset database to distinguish different calibration areas, and at the same time, the machine learning method is used to extract the image of the measurement target in the identification image positioned in the positioning step (S43) Feature: Match the learned image feature of the database image with the image feature of the measurement target, and obtain the measurement target matching the learned image feature corresponding to the measurement target of the current calibration area. In the case that there are multiple measurement items to be measured in a recognition image, the measurement targets are classified according to the learned image features that can distinguish different calibration areas, so that the measurement items in the recognition image will be recognized according to the measurement items of the anatomical structure Items are recognized. Among them, the method of extracting features by machine learning algorithms includes, but is not limited to, principal component analysis (Principle Components Analysis, PCA) method, linear discriminant analysis (Linear Discriminant Analysis, LDA) method, Harr feature extraction method, texture feature extraction method, etc. According to the learned image features that can distinguish different measurement items, the measurement items are classified to classify the measurement items in the recognition image. The classification discriminator used includes but not limited to the nearest neighbor classification (K nearest neighbor, KNN), support Vector machine (Suport vector maehine, SVM) random forest, neural network and other discriminators.

After the measurement target in the recognition image is obtained in the positioning step (S43), in the automatic measurement step (S44), the measurement target is measured to realize the automatic measurement process of the anatomical structure. The automatic measurement step (S44) The step of measuring a measurement target includes performing target fitting on the measurement target to obtain a fitting equation of the measurement target; and determining the measurement result of the measurement target through the fitting equation.

In one embodiment, the calibration result in the positioning step (S43) includes the ROI frame of the measurement target corresponding to the measurement item to be measured. In the automatic measurement step, the automatic measurement step (S44) measures the measurement target in the process of measuring the measurement target. Target fitting is performed on the measurement target within the frame area of the measurement target to obtain the fitted line, circle, or ellipse equation, and the measurement result is obtained by calculating the foregoing equation. Performing target fitting on the measurement target in the frame area of the measurement target can reduce the interference of other structures outside the frame area on the target fitting, and improve the accuracy of the measurement result.

In one embodiment, the calibration result in the positioning step (S43) includes a mask for accurately segmenting the measurement target corresponding to the measurement item. During the automatic measurement step (S44), the measurement target is measured. Target fitting is performed on the edge of the segmentation mask of the measurement target consistent with the calibration result in the recognition image, and the equations such as straight line, circle or ellipse are fitted, and the measurement result is obtained by calculating the foregoing equation. Target fitting is performed on the edge of the segmentation mask of the measurement target that is consistent with the calibration result in the recognition image, which can reduce the fitting error of the target fitting and improve the accuracy of the measurement result.

Although the exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above-described exemplary embodiments are merely exemplary, and are not intended to limit the scope of the present application thereto. Those of ordinary skill in the art can make various changes and modifications therein without departing from the scope and spirit of the present application. All these changes and modifications are intended to be included within the scope of the present application as required by the appended claims.

A person of ordinary skill in the art may realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

In the several embodiments provided in this application, it should be understood that the disclosed device and method may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another device, or some features can be ignored or not implemented.

In the instructions provided here, a lot of specific details are explained. However, it can be understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail, so as not to obscure the understanding of this specification.

Similarly, it should be understood that, in order to simplify this application and help understand one or more of the various aspects of the invention, in the description of the exemplary embodiments of this application, the various features of this application are sometimes grouped together into a single embodiment or figure. , Or in its description. However, the method of this application should not be interpreted as reflecting the intention that the claimed application requires more features than the features explicitly recorded in each claim. More precisely, as reflected in the corresponding claims, the point of the invention is that the corresponding technical problems can be solved with features that are less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are thus explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the application.

Those skilled in the art can understand that in addition to mutual exclusion between the features, any combination of all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and any method or device disclosed in this manner can be used. Processes or units are combined. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

In addition, those skilled in the art can understand that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments means that they are within the scope of the present application. Within and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

The various component embodiments of the present application may be implemented by hardware, or by software modules running on one or more processors, or by a combination of them. Those skilled in the art should understand that a microprocessor or a digital signal processor (Digital Signal Processor, DSP) may be used in practice to implement some or all of the functions of some modules according to the embodiments of the present application. This application can also be implemented as a device program (for example, a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the present application may be stored on a computer-readable medium, or may have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the application, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be constructed as a limitation to the claims. The application can be realized by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices may be embodied in the same hardware item. The use of the words first, second, and third, etc. do not indicate any order. These words can be interpreted as names.

The above are only specific implementations of this application or descriptions of specific implementations. The scope of protection of this application is not limited to this. Anyone familiar with the technical field within the technical scope disclosed in this application can easily Any change or replacement should be covered within the scope of protection of this application. The protection scope of this application shall be subject to the protection scope of the claims.

Claims

An automatic measurement method of anatomical structure, characterized in that it comprises:

In an image acquisition step, an ultrasound image is acquired, the ultrasound image is related to at least one anatomical structure of the biological tissue, and the at least one anatomical structure has at least one measurement item;

Steps for obtaining measurement items, obtaining the measurement items to be tested;

An image recognition step of identifying, from the ultrasound image, a recognition image containing a measurement target corresponding to the measurement item to be measured, and the recognition image is at least one of the ultrasound images;

A positioning step, positioning the measurement target in the recognition image;

In an automatic measurement step, the measurement target is measured.
An automatic measurement method of anatomical structure, characterized in that it comprises:

In an image acquisition step, an ultrasound image is acquired, the ultrasound image is related to at least one anatomical structure of the biological tissue, and the at least one anatomical structure has at least one measurement item;

Steps for obtaining measurement items, obtaining measurement items to be tested;

An image recognition step of identifying, from the ultrasound image, a recognition image containing a measurement target corresponding to the measurement item to be measured, wherein the recognition image is at least one of the ultrasound images;

In an automatic measurement step, the measurement target in the recognition image is measured.
The automatic measurement method according to claim 1 or 2, wherein the image recognition step comprises:

The image feature of the ultrasound image is compared with the image feature of the database image containing the measurement item to be measured in the preset database, and it is determined whether the image feature of the ultrasound image matches the image feature of the database image. When the image feature of the ultrasound image matches the image feature of the database image, it is determined that the ultrasound image is the recognition image.
The automatic measurement method according to claim 1 or 2, wherein the image recognition step comprises:

Classifying the ultrasound image according to the measurement items of the at least one anatomical structure of the biological tissue to obtain the ultrasound image containing the measurement target corresponding to the measurement item to be measured;

The ultrasound image containing the measurement target corresponding to the measurement item to be measured is selected from the classified ultrasound images.
The automatic measurement method according to claim 4, wherein the step of classifying the ultrasound image comprises:

The image feature of the ultrasound image is compared with the image feature of the database image in the preset database, wherein the database image contains at least one of the biological tissues of any one of the at least one of the anatomical structures For the measurement target corresponding to the measurement item, when the image feature of the ultrasound image matches the image feature of the database image, the ultrasound image includes the measurement target included in the database image.
The automatic measurement method according to claim 3, wherein at least two measurement items to be measured are obtained in the measurement item obtaining step, and the image recognition step further comprises:

The recognition image is classified to obtain the recognition image including the measurement target corresponding to each measurement item to be measured.
An automatic measurement method of anatomical structure, characterized in that it comprises:

An image acquisition step of acquiring at least two ultrasound images, the at least two ultrasound images being related to at least one anatomical structure of the biological tissue;

In an image recognition step, a recognition image containing a measurement target corresponding to a measurement item is recognized from the ultrasound image, wherein the measurement item is a measurement item to be measured, and the recognition image is at least one image in the ultrasound image ；

A positioning step, positioning the measurement target in the recognition image;

In an automatic measurement step, the measurement target is measured.
An automatic measurement method of anatomical structure, characterized in that it comprises:

An image acquisition step of acquiring at least two ultrasound images, the at least two ultrasound images being related to at least one anatomical structure of the biological tissue;

In an image recognition step, a recognition image containing a measurement target corresponding to a measurement item is recognized from the ultrasound image, wherein the measurement item is a measurement item to be measured, and the recognition image is at least one image in the ultrasound image ；

In an automatic measurement step, the measurement target is measured.
The automatic measurement method according to claim 7 or 8, wherein at least one anatomical structure of the biological tissue has at least one characteristic measurement item, and the image recognition step comprises:

Compare the image feature of the ultrasound image with the image feature of the database image containing the measurement target corresponding to any one of the at least one feature measurement item contained in the preset database, and determine the image feature of the current ultrasound image Whether it matches the image feature of the database image, when the image feature of the ultrasound image matches the image feature of the database image, it is determined that the current ultrasound image is the recognition image, wherein the recognition image contains The measurement target corresponds to the measurement target contained in the database image that matches the image feature of the recognition image.
The automatic measurement method according to claim 7 or 8, wherein any one or two of the at least one anatomical structure of the biological tissue has at least two characteristic measurement items, and the identification image contains at least two Each of the measurement items to be measured corresponds to the measurement target, and the image recognition step further includes:

The measurement targets in the recognition image are classified, so that the measurement targets in the recognition image correspond to the at least two feature measurement items in a one-to-one correspondence.
The automatic measurement method according to claim 1 or 7, wherein the measurement item positioning step comprises:

Compare and analyze the image feature of the recognition image with the image feature of the database image corresponding to the measurement item to be measured in the preset database, so as to locate the measurement item to be measured in the recognition image The measurement target is obtained, wherein the database image includes a calibration result corresponding to the measurement item to be measured, and the measurement target is an area consistent with the calibration result.
The automatic measurement method according to claim 1 or 7, wherein the identification image includes at least two measurement targets corresponding to the measurement items to be measured, and the measurement item positioning step further comprises:

The measurement target is classified, so that the measurement target corresponds to the at least two measurement items to be measured in a one-to-one correspondence.
The automatic measurement method according to claim 12, wherein the step of classifying the measurement target comprises:

The image feature of the measurement target is compared with the image feature of the calibration result corresponding to the measurement target corresponding to the measurement item to be measured in the database images contained in the preset database to determine the measurement target Whether the image feature of the measurement target matches the image feature of the calibration result, and when the image feature of the measurement target matches the image feature of the calibration result, it is determined that the measurement target corresponds to the measurement item to be measured.
The automatic measurement method according to claim 11, wherein the calibration result includes an ROI frame with the measurement target, and the positioning step includes:

Using a sliding window-based method to extract the image features in the sliding window, compare the image features in the sliding window with the image features of the calibration result, and judge the image features in the sliding window and the image of the calibration result Whether the features match, when the image feature in the sliding window matches the image feature of the calibration result, it is determined that the current sliding window is the measurement target.
The automatic measurement method according to claim 11, wherein the calibration result includes an ROI frame with the measurement target, and the measurement item positioning step includes:

According to the image feature of the calibration result in the database image containing the measurement target, frame regression processing is performed on the recognition image to obtain a frame area, and the frame area is the measurement target.
The automatic measurement method according to claim 11, wherein the calibration result includes a mask for accurately segmenting the measurement target, and the measurement item positioning step includes:

According to the image feature of the calibration result in the database image containing the measurement target, a semantic segmentation algorithm is used to identify the segmentation mask of the measurement target that is consistent with the calibration result in the recognition image.
The automatic measurement method according to claim 1 or 7, wherein the automatic measurement step comprises:

Performing target fitting on the measurement target to obtain a fitting equation of the measurement target;

The measurement result of the measurement target is determined by the fitting equation.
The automatic measurement method according to claim 2 or 8, wherein the automatic measurement step comprises:

Extracting the contour corresponding to the measurement target through an edge detection algorithm;

Fitting a contour corresponding to the measurement target to obtain a fitting equation corresponding to the measurement target;

The measurement result of the measurement target is determined by the fitting equation.
The automatic measurement method according to claim 1, 2, 7 or 8, characterized in that it further comprises:

The measurement result display step controls and displays the measurement result after the measurement target is measured.
An ultrasound imaging system, characterized in that it comprises:

Ultrasound probe, used to transmit ultrasonic waves to biological tissues and receive ultrasonic echoes to obtain ultrasonic echo signals;

A processor, configured to process the ultrasonic echo signal to obtain an ultrasonic image of the biological tissue;

A display for displaying the ultrasound image;

Memory, used to store executable program instructions;

The processor is further configured to execute the executable program instructions, so that the processor executes the automatic measurement method according to any one of claims 1-19.