JP6449457B2 - Image processing method and apparatus - Google Patents

Description

  The present invention relates to an ultrasonic apparatus, and more particularly to an image processing technique for detecting and providing standard cross-section information from ultrasonic data.

  An ultrasonic apparatus typified by an ultrasonic diagnostic apparatus has a characteristic that the inside of an object can be observed without destroying the object such as a human body or a structure. In particular, the human body does not require a surgical operation such as laparotomy, and is widely used in the medical field as a means for safely observing internal tissues.

  Among them, the heart is one of the objects in the ultrasonic diagnostic apparatus. For the heart cross section to be observed, several standard view guidelines have been defined so that anyone can observe it in a unified manner, and standard cross-sectional images based on these guidelines are defined by the ultrasound system. Is being made available at. The standard cross section of the heart includes a parasternal cross section, an apical view, a subcostal view, a suprasternal view, and the like.

  Conventionally, a method has been used in which an examiner operates an ultrasonic probe and images a heart section corresponding to these standard sections one by one. In recent years, for example, as shown in Patent Document 1, a three-dimensional three-dimensional ultrasonic heart image (hereinafter referred to as 3D volume data) is acquired using a special ultrasonic probe, and a plurality of 3D volume data are obtained from the 3D volume data. An ultrasonic apparatus that automatically obtains a standard cross-section of the above has been proposed.

JP2009-72593

  According to Patent Document 1, a 3D volume is provided with three types of discriminators that obtain a translation amount, a rotation amount, and an enlargement rate based on a target coordinate from which the features of the heart region have been learned in advance in order to obtain a target standard cross section. When a part of data is cut out and input, a translation method, a rotation amount, and an enlargement rate are obtained from a discriminator, and a processing method for obtaining a standard cross section calculated based on input data coordinates is described.

  However, 3D volume data is data with information on three axes (horizontal, vertical, and depth), and it is not possible to give this 3D volume data a translation amount, rotation amount, and enlargement ratio based on the coordinates of interest. This is equivalent to having the cube of 3 axes, that is, having 9 degrees of freedom, and it is practically difficult to obtain the target cross section directly.

  In addition, the above guidelines for obtaining a standard cross section define some recommendations such as capturing the position where the annulus diameter is maximum or capturing the position where the ventricular length is the longest. ing. However, in the method of Patent Document 1, even if the observation part of the target object, that is, the coordinates of the search part can be detected, the means for detecting the size and the length between the parts is not taken into consideration, so the annulus diameter is The maximum position or the like cannot be detected. That is, the method of Patent Document 1 cannot extract the recommended cross section defined by the guideline for obtaining the standard cross section.

  An object of the present invention is to solve the above-described problems and provide an image processing method, apparatus, and ultrasonic apparatus that can detect the size of a search part and the length between parts.

  In order to achieve the above object, in the present invention, an image processing method using a processing unit, the processing unit generates a first candidate cross section of an object using three-dimensional information of the object, From the second candidate cross-section and the first candidate cross-section obtained by transforming one candidate cross-section, a plurality of feature points that are search parts of the target object are detected, the plurality of feature points are classified into a predetermined feature point group, and the classified features Provided is an image processing method having a configuration in which a center point is obtained for each point group and cross sections passing through a plurality of center points are used as observation cross sections of an object.

  In order to achieve the above object, according to the present invention, an image processing apparatus including a processing unit generates a first candidate cross section of a target object using three-dimensional information of the target object. Then, from the second candidate cross section obtained by deforming the first candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the object are detected, and the plurality of feature points are classified into a predetermined feature point group, Provided is an image processing apparatus having a configuration in which a center point is obtained for each point group, and cross sections passing through the obtained plurality of center points are used as observation cross sections of an object.

  Furthermore, in order to achieve the above object, in the present invention, the ultrasonic device is a three-dimensional information of a target object based on a received signal from an ultrasonic probe that transmits / receives ultrasonic waves to / from the target object. And a display unit that displays a cross-sectional image of the target object generated by the processing unit, the processing unit generates a first candidate cross-section of the target object using the three-dimensional information, From the second candidate cross section and the first candidate cross section obtained by deforming the first candidate cross section, a plurality of feature points that are search parts of the target object are detected, the plurality of feature points are classified into a predetermined feature point group, and the feature point group Provided is an ultrasonic device configured to obtain a center point every time, and to control a cross section passing through the obtained plurality of center points as an observation cross section of an object and display an image of the cross section for observation on a display unit.

  According to the present invention, it is possible to acquire a cross-sectional image at the most appropriate position by detecting the size of the search part and the length between parts.

1 is a block diagram of an ultrasonic apparatus that acquires a standard cross-sectional image according to Embodiment 1. FIG. FIG. 3 is a block diagram showing internal processing of a 3D search unit according to the first embodiment. FIG. 3 is a diagram illustrating an example of a feature classifier of the 3D search unit according to the first embodiment. The figure which shows an example of 2D cross-section cut-out of the 3D search part based on Example 1. FIG. The figure which shows an example which reconfigure | reconstructs the three-dimensional shape of the search part based on Example 1. FIG. The figure which shows an example of the standard cross sections A4C, A2C, and PLX based on Example 1. FIG. The figure which shows an example of the standard cross section PSX MV level, PM level, and Apex level based on Example 1. FIG. FIG. 3 is a flowchart illustrating an example of 2D identification processing according to the first embodiment. FIG. 6 is a diagram illustrating an example of feature point group generation according to the first embodiment. FIG. 3 is a diagram illustrating an example of feature point center detection according to the first embodiment. FIG. 3 is a diagram illustrating an example of standard section extraction according to the first embodiment. FIG. 10 is a diagram illustrating an example of apex center detection processing according to the second embodiment. The flowchart figure which shows the whole flow of standard cross-section extraction based on each Example.

  Hereinafter, various embodiments of the present invention will be described with reference to the drawings. The present invention is not limited to the human body, particularly the heart, but can be applied to other structures and the like. However, in the following description of the embodiments, the human body, particularly the heart will be described as an example. Note that the above-mentioned guidelines recommended for obtaining the standard cross-section of the heart are examples in the case of acquiring a 4 apical chamber (A4C) cross-section. In addition, the standard cross sections of the apical 2 chamber (A2C: Apical 2 Chamber), the left sternal long axis (PLX: Parasternal Long Axis), and the left sternal short axis (PSX) are also observed. Recommendations are defined. The mutual positional relationship between the standard cross sections of A4C, A2C, PSX and PLX will be sequentially described in the description of the following examples.

  Example 1 is an image processing method and apparatus using a processing unit, and the processing unit generates a first candidate cross section of the target object using the three-dimensional information of the target object, and the first candidate cross section is generated. From the deformed second candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the target object are detected, and the plurality of feature points are classified into a predetermined feature point group, and for each classified feature point group It is the Example of the image processing method of the structure which calculates | requires a center point, and makes the cross section which passes through several center points into the cross section for observation of a target object, and its apparatus.

  In addition, the first embodiment is an ultrasonic apparatus, and a processing unit that processes three-dimensional information of a target object based on a reception signal from an ultrasonic probe that transmits and receives ultrasonic waves to the target object, and a processing A display unit that displays a cross-sectional image of the target object generated by the unit, and the processing unit generates a first candidate cross-section of the target object using the three-dimensional information, and deforms the first candidate cross-section From the second candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the target object are detected, the plurality of feature points are classified into a predetermined feature point group, and a center point is obtained and obtained for each feature point group. This is an embodiment of an ultrasonic apparatus configured to control a cross section passing through a plurality of center points as an observation cross section of an object and display an image of the observation cross section on a display unit.

  FIG. 1 is a block diagram of a configuration example of an ultrasonic apparatus according to the first embodiment that acquires a standard cross-sectional image of an object. The basic configuration of this ultrasonic apparatus is a configuration that is commonly used in other embodiments.

  In the figure, an ultrasonic device 100 is a device that acquires and displays an image of the inside of a target object by bringing an ultrasonic probe 110 into contact with the target object. An ultrasonic probe is a sensor device incorporating a transducer that generates or receives ultrasonic waves. The principle of ultrasonic generation is that a high-frequency pulse voltage is applied to the piezoelectric element material, and ultrasonic waves are generated by the distortion (vibration) of the element. Ultrasonic waves have the property of passing or reflecting by objects existing in the propagation direction. Also, the propagation speed in the object varies depending on the type of the object. The ultrasonic reflected signal is received again by the ultrasonic probe 110, and the inside of the object is visualized by converting the received signal intensity into an image and displaying it. The ultrasonic probe 110 used in the ultrasonic apparatus 100 of this embodiment includes a transducer, an ultrasonic transmission / reception circuit, a phasing addition circuit for phasing and adding received echo signals, an analog digital (A / D) converter, and the like. Is a device that can acquire 3D volume data by ultrasonic waves, that is, 3D information, and the ultrasonic apparatus 100 receives 3D volume data as acquired 3D information.

  The data input unit 101 is an input interface that receives 3D volume data from the ultrasound probe 110. The 3D volume data that is the three-dimensional information received by the data input unit 101 is temporarily stored in the memory 102. The three-dimensional information stored in the memory 102 is processed by a 3D search unit 103, a 2D image generation unit 104, a 2D identification unit 105, a cross-section position adjustment unit 106, a standard cross-section extraction unit 107, and the like, which will be sequentially described below. Each function of the 3D search unit 103 to the standard cross section extraction unit 107 can be realized by executing a program of a central processing unit (CPU) 109 which is a processing unit described later. In addition, the ultrasonic apparatus 100 includes a cross-section output unit 108 that outputs cross-sectional image data, and an input interface (I / F) 112 for inputting a user operation. The input I / F 112 is acquired and acquired by a user operation 113. Standard section setting and display method to be displayed are input to the device.

  Now, the 3D search unit 103 reads 3D volume data that is three-dimensional information stored in the memory 102, and a search part that is a feature point based on the observation position by a plurality of feature classifiers provided in the 3D search unit 103 Output vector information to.

  The 2D image generation unit 104 refers to two or more pieces of vector information output by the 3D search unit 103, determines 2D cross-sectional coordinates in the stored 3D volume data, and includes pixel data on the cross-section. Generate a 2D image. In this 2D image generation, a plurality of 2D images are obtained while changing the observation position so as to three-dimensionally include the target search portion. The number of 2D images to be generated may be arbitrarily set within a range in which the shape of the search site can be observed three-dimensionally.

  The 2D identification unit 105 receives the 2D image generated by the 2D image generation unit 104 and extracts a desired number of feature point groups.

  The cross-sectional position adjustment unit 106 obtains the center coordinates of the site to be observed from the plurality of 2D images output from the 2D image generation unit 104, that is, detects the feature point center of the search site.

  The standard cross section extraction unit 107 determines the central coordinates of the plurality of search parts obtained by the cross section position adjustment unit 106, that is, a cross section connecting the feature point centers as a standard cross section, and extracts a 2D image of this cross section.

  The CPU 109 serving as a processing unit executes each functional process of the 3D search unit 103 to the standard cross section extraction unit 107 described above according to a procedure programmed in advance and stored in the apparatus.

  The image display device / recording medium 111 is a monitor or recording medium connected to the ultrasonic device 100, and the user uses the standard cross section extracted by the ultrasonic device 100 using the cross-sectional image data output from the cross-section output unit 108. Is displayed on a monitor or stored in a recording medium in accordance with the display method designated by.

  FIG. 13 is a flowchart showing the entire flow of standard section extraction by the ultrasonic apparatus of the first embodiment described above. First, the data input unit 101 inputs 3D volume data of an object from the ultrasound probe 110 (1301), the 3D search unit 103 executes a 3D search (1302), and the 2D image generation unit 104 An initial cross section of the search site as a candidate cross section is extracted (1303).

  Subsequently, the 2D identification unit 105 performs a cross-sectional movement (1306) with deformations such as left and right, up and down, front and rear, and rotation from the initial cross section that is the first candidate cross section to obtain a plurality of second candidate cross sections, Using the first and second candidate sections, a plurality of feature point groups are extracted by 2D identification for detecting a search site (1304). It is determined whether or not a desired feature point group, for example, a predetermined number of feature points has been acquired (1305), and the generation of the feature point group is terminated when the acquisition is completed. Then, the cross-section position adjustment unit 106 detects the feature center, that is, the center of the observation site, from the plurality of feature point groups extracted by the 2D identification unit 105 (1307), and the standard cross-section extraction unit 107 passes through each feature center. The 2D plane is determined as a standard cross section (1308) and output to a monitor, an external recording medium or the like (1309), and the standard cross section extraction flow is completed.

  Hereinafter, each function of the 3D search unit 103 to the standard cross-section extraction unit 107 will be sequentially described using specific examples. FIG. 2 is a block diagram showing internal processing of the 3D search unit 103. As shown in FIG. The 3D search unit 103 detects a feature point of the search site from the 3D volume data 200 stored in the memory 102, and outputs a 2D cross-sectional image of the search site based on a plurality of feature point coordinates.

  As a method for detecting the feature point of the search site, for example, there is a method using Random Forest. Random Forest is one of the machine learning methods that uses the advantage of randomness to efficiently learn a large number of decision trees. For example, by preparing a decision tree that performs detailed judgments regarding the direction and distance between various feature points existing in the known 3D volume data 200, and combining them, the unknown 3D volume data 200 can be Similarly, a feature classifier 201 capable of obtaining the direction and distance between arbitrary feature points can be generated. Thereby, it is possible to know the direction and distance to the search site with the target site data input from the unknown 3D volume data 200 as a base point.

  The 3D search unit 103 takes the following processing procedure. First, attention site data selected from unknown 3D volume data 200 is input to the feature classifier 201. The feature classifier 201 repeats the condition determination regarding the direction and the distance to the search site with the target site as a base point by using a plurality of decision trees, and calculates appropriate vector information.

  The vector information integration 202 obtains vector information for each of the search parts for a plurality of target parts selected from the 3D volume data 200, and then aggregates the plurality of vector information and selects one cross section. To integrate vector information. The 2D cross-section coordinate calculation 203 calculates plane coordinates necessary for 2D cross-section extraction from the integrated vector information. In the 2D section cutout 204, a 2D section image 205 is extracted from the plane coordinates based on the calculated plane coordinates.

  In the 3D search unit 103, a method using Random Forest has been described as a method for obtaining the feature point coordinates of the search site, but other methods may be used. For example, a feature point of a search part may be detected by pattern matching between an input 3D image and a reference 3D image prepared in advance.

  FIG. 3 is a schematic diagram illustrating an example of the feature classifier 201 of the 3D search unit 103 of the present embodiment. As described above, the 3D search unit 103 using the Random Forest prepares a decision tree for performing a detailed determination regarding distance, and by combining a plurality of them, the unknown sampling 3D volume data (Vn) 301 is also obtained. The direction and distance between arbitrary feature points can be obtained. The circles in the figure are the decision tree 302, and the squares are the vector information 303 for the search feature points that are finally determined. Each decision tree 302 holds determination conditions derived from learning regarding the direction and distance between the input part in the 3D volume data of the learning sample and the target part existing at a predetermined position. When the unknown sampling 3D volume data 301 is input, the search is finally made to the position of the most appropriate search part by sequentially passing through a decision tree 302 having the same direction and distance relation between the input part image and the target part. Vector information 303 to the feature point can be reached.

  FIG. 4 is a diagram illustrating an example of 2D section cutout of the 3D search unit of the present embodiment. As shown in the figure, a 2D cross section including a feature point indicating the position of the search site can be cut out based on a plurality of vector information obtained by the 3D search unit 103. As for vector information, if there is information on three or more points, a 2D planar image can be cut out.

  Therefore, for example, there are three points of interest, v1 (401), v2 (402) and v3 (403), and for each of the feature points t1 (407), t2 (408) and t3 (409) as search parts Assume that 3 points have been obtained. At this time, a line d1 (404) connecting the attention site v1 (401) and the feature point t1 (407), a line d2 (405) connecting the attention site v2 (402) and the feature point t2 (408), and attention A line d3 (406) connecting the part v3 (403) and the feature point t2 (409) leads to a plane 410 that has a normal vector relationship. The plane 410 from which the normal vector relationship is obtained becomes a 2D cross-sectional image to be cut out as the first candidate cross-section.

  FIG. 5 uses a 2D cross-sectional image that is the first candidate cross-section obtained by the 3D search unit of the present embodiment, obtains a plurality of feature point groups, and reconstructs the three-dimensional shape of the search site from the plurality of feature points. It is a figure which shows an example. In one 2D cross-sectional image 505 (410) acquired by the 3D search unit 103, there is only one feature point 504 of one search site. On the other hand, taking the tricuspid valve of the heart as an example, the actual shape of the tricuspid valve is a substantially circular (or elliptical) closed shape with some deformation. Since there is only one feature point 504 of the search site corresponding to the target site v1 (401) in the 2D cross-section image 505 (410) that is one first candidate cross-section, the feature point is a tricuspid valve. There is no guarantee that it is central.

  Therefore, in the apparatus according to the present embodiment, the 2D cross-sectional image 505 (410) is obtained with reference to the horizontal axis 508 or the vertical axis 509 of the two-dimensional plane of the acquired 2D cross-sectional image 505 (410) as the first candidate cross section. Thus, a plurality of 2D cross-sectional images, which are second candidate cross-sections, are acquired while arbitrarily giving deformation such as left-right movement, vertical movement, back-and-forth movement 506, 507, and rotational movement 501, 502. A feature point group in which a search site can be observed in three dimensions by detecting feature points from each of a plurality of 2D cross-sectional images composed of one first candidate cross-section and a plurality of second candidate cross-sections. 503 can be acquired. Then, using the obtained feature point group 503, it is possible to calculate the center of the search site, for example, the center of the circle of the tricuspid valve or the mitral valve, and to obtain a standard cross section passing through the center of the search site become.

  In the A4C feature point group extraction, which is the four-chamber apex image, the center of the line connecting the feature points of the apex obtained by the first feature point detection of the 3D search unit 103 and the feature points of the mitral and tricuspid valves A feature line group extraction step is omitted by acquiring a plurality of 2D cross-sectional images while arbitrarily deforming such as horizontal movement, vertical movement, back-and-forth movement, and rotational movement with a single line passing through a point as a base axis. May be. In A4C, which is an image of the apex of four cavities, the shape characteristics of the apex, mitral valve, and tricuspid valve can be obtained by extracting the feature points by only moving the basic axis as described above, thereby obtaining a three-dimensional feature point group for each part. is there. Similarly, regarding the feature point group extraction of other standard sections, the movement of the base axis may be limited in accordance with each shape characteristic.

  By having such a configuration of the present embodiment, the position where the annulus diameter defined in the guidelines for obtaining the standard cross section is maximized is captured, or the position where the ventricular length is the longest is captured. It is possible to guarantee such recommendations. Hereinafter, the reason will be described with reference to FIGS.

  FIG. 6 is a schematic diagram showing an example of standard cross sections A4C, PLX, and A2C. The upper part of FIG. 6 shows the positions of the standard sections A4C601, PLX602, and A2C603 with respect to the heart, and the lower part of FIG. 6 shows an example of each standard section. Each feature point of the standard section at the bottom of FIG. 6 will be described. The apex 604 is the apex of the heart, the mitral valve 605 is the valve between the left ventricle and the left atrium, the tricuspid valve 606 is the valve between the right ventricle and the right atrium, and the aortic valve 607 is between the left ventricle and the aorta There is a valve. The connection between each valve and the myocardium is called an annulus.

  A4C601 is a cross section through the mitral valve, tricuspid valve and apex, PLX602 is a cross section through the mitral valve, aortic valve and apex, A2C603 is a cross section through the mitral valve and the apex, left ventricle and left Only the atrium is included in the image. After satisfying each feature, the recommended cross section is the cross section with the maximum diameter of the valve and the longest distance between the mitral valve and the apex.

  FIG. 7 is a schematic diagram showing an example of the MV level, PM level, and Apex level of the standard cross section PSX. 7 shows the positions of the standard cross-section PSX MV level 701, PM level 702, and Apex level 703 with respect to the heart, and the lower part of FIG. 7 shows an example of cross-sectional images of the standard cross-section levels. The standard cross section PSX is a cross section parallel to the mitral annulus and is classified into three stages according to the position. PSX MV level 701 is the recommended cross section near the mitral valve, PSX PM level 702 is the vicinity of the connection between the papillary muscle and myocardium, and PSX Apex level 703 is the recommended cross section near the apex.

  FIG. 8 is a flowchart showing an example of feature point group acquisition by 2D identification processing by the 2D classifier in the present embodiment. In the cross-sectional image input (801), the cross-sectional image generated by the 2D image generation unit 104 is input. The standard cross section identification (802) classifies the input cross section image into standard cross sections A4C, PLX, A2C, PSX, and undecidable (803). This standard section identification method uses, for example, a neural network or a support vector machine.

  The feature point extraction (804) extracts feature points in the cross-sectional image according to the classified cross-section type. At this time, the type of feature point to be extracted is selected according to the result of the standard cross section identification. For A4C, the apex, mitral annulus, tricuspid annulus, for PLX, extract the apex, mitral annulus, aortic annulus, for A2C, extract the apex and mitral annulus. As the feature point extraction method, for example, Random Forest, ORB (Oriented FAST and Rotated BRIEF), pattern matching, or the like is used.

  The feature point coordinate recording (805) records the obtained 3D coordinates of the feature point group in the memory. For example, when the 2D image generation unit generates the cross-sectional image, the 3D coordinate of each pixel is recorded in the memory, and the corresponding pixel is converted from the 2D coordinate on the cross-sectional image. Use a method to obtain 3D coordinates.

  FIG. 9 is a schematic diagram illustrating an example of feature point group generation in the 2D identification unit. FIG. 9 shows generation of feature point groups of the mitral annulus as an example, but feature point groups can be similarly generated for the apex, tricuspid valve, and aortic valve. The upper part of FIG. 9 shows the positions of a plurality of cross-sectional images translated in parallel with the heart, the middle part of FIG. 9 schematically shows a plurality of cross-sectional images, and the lower part of FIG. 9 schematically shows a feature point group on 3D coordinates.

  As shown in the middle part of FIG. 9 by extracting characteristic points (804), a mitral annulus 903 as a search site is obtained from the cross-sectional image 901. A cross-sectional image 902 is an example of a cross-sectional image obtained by translating the cross-sectional image 901 in the normal direction. Similarly, the mitral annulus 904 is extracted from the cross-sectional image 902. As shown in the lower part of FIG. 9, the feature point group 905 is generated by converting the coordinates of the mitral annulus into 3D coordinates and plotting them on 3D data.

  First, as described with reference to FIG. 13, generation of a feature point group ends when acquisition of a desired feature point group, for example, a predetermined number of feature points is completed. In other words, for example, the number of feature points required for each feature is determined in advance, and a method for determining whether the number of feature points exceeds this, or the number of cross-sectional images from which features are extracted, The determination is made by a method of determining whether or not the number of cross sections for extracting features exceeds this, or a combination of both.

  FIG. 10 is a diagram for explaining an example of feature point center detection by the cross-section position adjusting unit in the present embodiment. The upper part of the figure schematically shows the three feature point groups as search parts, A4C, and PSX MV level 1001, and the middle and lower parts of the figure are diagrams for explaining the feature point center detection of the feature point group. Shown schematically.

As the feature point center of the annulus, for example, a feature point group is projected on an arbitrary axis, and the middle of the two most distant points is used. This ensures that the position where the annulus diameter defined in the guidelines for standard cross-section acquisition is maximized is captured. As shown on the left side of the middle part of the figure, the annulus one-dimensional feature point group 1002 is an example in which the feature point group is projected on the x-axis. The feature point 1003 is a feature point that is x _min , and the 3D coordinates are (x ₁ , y ₁ , z ₁ ). The feature point 1004 is a feature point that is x _max , and the 3D coordinate is (x ₂ , y ₂ , z ₂ ). Since feature points 1003 and 1004 are both ends of the annulus diameter, they are 3D coordinates ((x ₂ -x ₁ ) / 2, (y ₂ -y ₁ ) / 2, (z ₂ -z ₁ ) / 2) The midpoint 1005 is the feature point center. The same applies to the tricuspid valve center and the aortic valve center.

  As shown in the middle right of FIG. 10, as the apex center where the ventricular length is the longest, for example, the feature point group of the apex is projected around the distance from the mitral valve center, and the feature point farthest from the mitral valve is projected. Use. The apex one-dimensional feature point group 1006 is an example in which the feature point group is projected onto the distance axis from the center of the mitral valve. The feature point 1007 is the feature point farthest from the mitral valve and is selected as the apex center. Thereby, it can be ensured that the position where the ventricular length defined in the guideline for obtaining the standard section is the longest is captured. Similarly, also in the case of the cross-sectional image of the PSX MV level 1001 in the lower part of the figure, the feature point centers 1008 and 1009 can be obtained using the middle of the two most distant points of each of the two feature point groups.

  FIG. 11 is a schematic diagram illustrating an example of standard section extraction by the standard section extraction unit of the present embodiment. In the above-described processing, as shown in the upper and middle stages of the figure, the apex center 1107, the mitral valve center 1108, the tricuspid valve center 1109, and the aortic valve center 1110 are obtained from the cross-sectional images 1101, 1102, and 1103, respectively. And

  Cross-sectional image 1101 corresponding to A4C passes through apex center 1107, mitral valve center 1108, and tricuspid valve center 1109, and cross-sectional image 1102 corresponding to PLX passes through apex center 1107, mitral valve center 1108, and aortic valve center 1110 Extract the cross section. The cross-sectional image 1103 corresponding to A2C needs to include only the left ventricle and the left atrium. Therefore, a cross section passing through the apex center 1107 and the mitral valve center 1108 and the apex-mitral valve axis 1111 connecting the apex and the mitral valve and the tricuspid valve shown in the lower part of the figure is extracted as A2C.

  As shown in the lower part of the figure, PSX extracts a plane parallel to the mitral valve annulus by translating it in the direction of the apex-mitral valve axis 1111. For the plane parallel to the mitral annulus, for example, three feature points on the annulus diameter are obtained by the method described in FIG. 10, and a cross section passing through these is used. The amount of translation can be determined by setting the ratio in advance, for example, mitral valve center is 0, apex center is 1, MV level is 1/6, PM level is 3/6, Apex level is 5/6. Extraction can obtain cross sections 1106, 1105, 1104.

  As described in detail above, by detecting the size of the observation region and the length between the observation regions by the apparatus and method of the present embodiment, a cross-sectional image at the most appropriate position can be acquired, and the recommended criteria are met. Standard section can be extracted.

  Example 2 is an example of an image processing method, an apparatus thereof, and an ultrasonic apparatus, as in Example 1, and is an example in which the processing of the cross-sectional position adjustment unit of Example 1 is changed.

  In the second embodiment, in the feature point center detection process in the cross-section position adjustment unit 106 of the first embodiment, when the apex is unclear, the feature point center is detected by, for example, extracting the myocardium near the apex as a feature point group. This is an embodiment. Since all that is necessary is to perform the same processing as that in the first embodiment, a duplicate description is omitted here.

  FIG. 12 is a schematic diagram illustrating an example of apex center detection processing according to the present embodiment. As shown in the upper part of the figure, the apex is unclear, and unlike the case of Example 1, the feature points of the apex cannot be detected. Instead, as shown on the cross section 1201, the myocardium near the apex is the feature point group. , The center of the myocardial feature point group is obtained by the method described with reference to FIG. Thus, according to the present embodiment, even if the apex is unclear, the standard cross section can be extracted according to the recommended standard as in the first embodiment.

  According to the present invention described in detail above, the three-dimensional shape of the search part is reconstructed from the feature point group of the search part detected by the 3D search for the 3D volume data, and the size of the search part and the length between the observation parts are detected. It becomes possible to do.

  According to the present invention, in the image processing method and apparatus for acquiring a standard cross section of the heart from 3D volume data, the three-dimensional shape of the search part is reconstructed from the feature point group of the search part, and the size and search of the search part By providing the means and function for detecting the length between the parts, it is possible to acquire a cross-sectional image at the most appropriate position in accordance with the recommendations of the guideline regarding standard cross-section acquisition.

  Furthermore, according to the present invention, recommendations such as capturing the position where the annulus diameter defined in the guidelines for obtaining a standard cross section is the maximum, or capturing the position where the ventricular length is the longest, are provided. It is possible to provide a satisfactory ultrasonic apparatus.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for better understanding of the present invention, and are not necessarily limited to those having all the configurations described. For example, the target body has been described by taking the heart of a human body as an example, but it may be another organ of the human body, or a structure other than the human body can be used as the target body. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment.

  Further, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment. For example, when the target object is the same, the above-described ultrasonic apparatus superimposes the control parameters relating to the coordinates of the first candidate cross section, the rotation of the first candidate cross section when generating the second candidate cross section, the back and forth movement, and the left and right movement. By storing in the memory of the sonic device and reusing it, the processing steps of the first candidate cross-section extraction and the second candidate cross-section extraction can be omitted, and the processing time until the observation cross-section determination and extraction can be shortened. it can.

  Furthermore, although each of the above-described configurations, functions, processing units, and the like has been described as an example of creating and realizing a program for a CPU that realizes part or all of them, some or all of them are, for example, integrated circuits. It goes without saying that it may be realized by hardware by designing or the like.

100 ultrasonic equipment
101 Data input section
102 memory
103 3D search unit
104 2D image generator
105 2D identification part
106 Section position adjustment section
107 Standard section extractor
108 Section output section
109 CPU
112 Input I / F
201 Feature classifier
202 Vector information integration
203 2D section coordinate calculation
204 2D cut out
205, 410, 505 2D cross-sectional image
301 Sampling 3D volume data (Vn)
302 decision tree
303 vector information
503, 905 feature points
1107 Center of apex
1108 Mitral valve center
1109 Tricuspid valve center
1110 Aortic valve center
1111 Apex-mitral valve stem

Claims (15)

An image processing method using a processing unit,
The processor is
Using the 3D information of the object, generating a first candidate cross section of the object;
From the second candidate cross section obtained by deforming the first candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the object are detected,
Classifying the plurality of feature points into a predetermined feature point group;
A center point is obtained for each feature point group,
A cross section that passes through the plurality of obtained center points is an observation cross section of the object.
An image processing method. An image processing method according to claim 1,
The processor is
Obtaining one two-dimensional plane in which a normal vector relationship is established with each of three points of interest in the three-dimensional information, and determining the obtained two-dimensional plane as the first candidate cross section;
An image processing method. An image processing method according to claim 1,
The processor is
The second candidate cross section obtained by deforming the first candidate cross section is rotated, vertically moved, moved left and right, back and forth, with respect to a horizontal axis and a vertical axis passing through the center coordinates of the first candidate cross section. Generate and perform any movement
An image processing method. An image processing method according to claim 1,
The processor is
Find the position where the diameter of each shape of the feature point group mapped on the three-dimensional information is maximum, determine the center position of the determined maximum diameter as the center point,
An image processing method. An image processing method according to claim 1,
The processor is
For each object, the coordinates of the first candidate cross section and the control parameters related to the deformation when generating the second candidate cross section are stored.
An image processing method. An image processing apparatus including a processing unit,
The processor is
Using the 3D information of the object, generating a first candidate cross section of the object;
From the second candidate cross section obtained by deforming the first candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the object are detected,
Classifying the plurality of feature points into a predetermined feature point group;
A center point is obtained for each feature point group,
A cross section that passes through the plurality of obtained center points is an observation cross section of the object.
An image processing apparatus. An image processing apparatus according to claim 6,
The processor is
Obtaining one two-dimensional plane in which a normal vector relationship is established with each of three points of interest in the three-dimensional information, and determining the obtained two-dimensional plane as the first candidate cross section;
An image processing apparatus. An image processing apparatus according to claim 6,
The processor is
The second candidate cross section obtained by deforming the first candidate cross section is rotated, vertically moved, moved left and right, back and forth, with respect to a horizontal axis and a vertical axis passing through the center coordinates of the first candidate cross section. Generate and perform any movement
An image processing apparatus. An image processing apparatus according to claim 6,
The processor is
Find the position where the diameter of each shape of the feature point group mapped on the three-dimensional information is maximum, determine the center position of the determined maximum diameter as the center point,
An image processing apparatus. An image processing apparatus according to claim 6,
A storage unit;
The processor is
For each of the objects, the coordinates of the first candidate cross section and the control parameters related to the deformation when generating the second candidate cross section are stored in the storage unit.
An image processing apparatus. An ultrasonic device,
A processing unit for processing the three-dimensional information of the target object based on a reception signal from an ultrasonic probe that transmits and receives ultrasonic waves to the target object,
A display unit that displays a cross-sectional image of the object generated by the processing unit,
The processor is
Generating a first candidate cross section of the object using the three-dimensional information;
From the second candidate cross section obtained by deforming the first candidate cross section and the first candidate cross section, a plurality of feature points that are search parts of the object are detected,
Classifying the plurality of feature points into a predetermined feature point group;
A center point is obtained for each feature point group,
A cross section that passes through the plurality of obtained center points is an observation cross section of the object,
Control to display an image of the observation cross section on the display unit;
An ultrasonic device characterized by that. An ultrasonic device according to claim 11,
The processor is
Obtaining one two-dimensional plane in which a normal vector relationship is established with each of three points of interest in the three-dimensional information, and determining the obtained two-dimensional plane as the first candidate cross section;
An ultrasonic device characterized by that. An ultrasonic device according to claim 11,
The processor is
The second candidate cross section obtained by deforming the first candidate cross section is rotated, vertically moved, moved left and right, back and forth, with respect to a horizontal axis and a vertical axis passing through the center coordinates of the first candidate cross section. Generate and perform any movement
An ultrasonic device characterized by that. An ultrasonic device according to claim 13,
The object is a living heart,
The processor is
Rotation, vertical movement of the first candidate cross section, with a line passing through two points of the center point of the apex of the first candidate cross section and the center point of the line connecting the center points of the mitral and tricuspid valves as the base axis, Execute left and right movement, back and forth arbitrarily, and detect a plurality of the feature points,
An ultrasonic device characterized by that. An ultrasonic device according to claim 11,
The processor is
Find the position where the diameter of each shape of the feature point group mapped on the three-dimensional information is maximum, determine the center position of the determined maximum diameter as the center point,
An ultrasonic device characterized by that.

Images