JP7696641B2 - Ultrasound image search device and program - Google Patents

Description

The present invention relates to a device and a program for searching for ultrasound images suitable for diagnosis in ultrasound examinations.

In recent years, cardiac ultrasound examinations (echocardiograms), which non-invasively and highly accurately support doctors in diagnosing heart diseases, which account for the highest mortality rates in Japan, have been attracting attention. Echocardiograms are performed to evaluate the morphology and movement of the subject's heart by applying ultrasound to the subject's heart and converting the reflected sound waves into electrical signals to depict them as images. In order to make an appropriate diagnosis, echocardiograms require advanced techniques to accurately and clearly depict two-dimensional ultrasound images of the heart, so that examiners such as doctors and technicians who perform the examination are required to have appropriate knowledge and experience. However, there are only a limited number of hospitals that employ experienced examiners, and many medical institutions currently lack examiners who can perform echocardiograms. Furthermore, in echocardiograms, the ultrasound probe held by the examiner is applied to the body surface of the subject, such as a patient, at various angles, and the subject is forced to maintain an unnatural posture, so the burden on the subject increases as the examination time increases. Therefore, the inventors have developed an ultrasound examination robot that automates echocardiography, which has previously required an examiner to manually operate an ultrasound probe to search for appropriate ultrasound images (see Patent Document 1). In this ultrasound examination robot, the ultrasound probe, which is positioned relative to the examinee's chest, is automatically moved along the surface of the examinee's chest while the examinee is rotated laterally and forwards and backwards in a seated position. This eliminates the need for the examiner to perform the previous complex manual operation of the ultrasound probe, and allows the examinee to undergo an echocardiography with reduced examination burden.

In the ultrasound examination robot, when controlling the operation of the ultrasound probe by the robot, it is necessary to specify the position and posture of the ultrasound probe that will provide an appropriate ultrasound image of the examination plane for determining the presence or absence of heart disease by echocardiography. Here, as examination planes for determining the presence or absence of heart disease, it is necessary to depict basic planes called the left parasternal left ventricular long axis plane, the left parasternal left ventricular short axis plane, and the apical four-chamber plane. Of these, for the left parasternal left ventricular long axis plane, the inventors have already proposed an algorithm that automatically specifies the position information and posture information of the ultrasound probe that can obtain the desired ultrasound image (see Patent Document 2).

JP 2023-70607 A JP 2023-70601 A

Through further research by the inventors, for the ultrasound image of the apical four-chamber section among the basic sections, an algorithm was created that uses information identified when acquiring an ultrasound image of the left ventricular long axis section using the algorithm of Patent Document 2 to identify the position and orientation of the ultrasound probe that will obtain an ultrasound image of the apical four-chamber section that is appropriate for diagnosis, with efficient operation of the ultrasound probe.

The present invention was devised based on the research of the inventors, and its purpose is to provide an ultrasound image search device and a program for the same that can easily obtain ultrasound images of the apical four-chamber section suitable for diagnosis with efficient operation of an ultrasound probe.

To achieve the above object, the present invention is primarily an apparatus for searching for a target image depicting a desired examination cross section for echocardiography from a plurality of cardiac ultrasound images obtained while changing the position and posture of an ultrasound probe in contact with the chest surface of a subject, and is provided with an image analysis unit that estimates the proper positioning consisting of the desired position and desired posture of the ultrasound probe at which the target image is obtained by image analysis of each of the ultrasound images, and extracts the ultrasound image obtained at the proper positioning as the target image, and the image analysis unit performs image analysis on each of the ultrasound images to obtain detection information including an inference value representing the probability of the existence of a specific part in the heart that needs to be depicted in the target image, and is provided with a probe placement identification unit that identifies the proper placement of the ultrasound probe on the surface of the apex of the heart of the subject.

In the claims and the present specification, the directions of the position and posture of the ultrasound probe are the directions shown in FIG. 1 unless otherwise specified. That is, the "x-axis direction" in the coordinate system of three orthogonal axes representing the position of the ultrasound probe P is the head-tail direction of the subject, the "y-axis direction" is the lateral (left-right) direction of the subject, and the "z-axis direction" is the normal direction of the subject's body surface. In addition, when the center of the lower end of the ultrasound probe P in the direction shown in the figure is the center of rotation, the "roll direction" representing the posture of the ultrasound probe P is the rotation direction φ around the axis (the x-axis) perpendicular to the beam scanning plane F, the "pitch direction" is the rotation direction θ in the direction (around the y-axis) that swings the beam scanning plane F back and forth, and the "yaw direction" is the rotation direction ψ around the z-axis.

In the present invention, when searching for a target image depicting the apical four-chamber section for echocardiography from multiple cardiac ultrasound images obtained by the operation of a specified ultrasound probe, the appropriate positioning of the ultrasound probe on the apical surface to obtain the target image is automatically identified. In other words, the appropriate positioning is automatically identified by analyzing the ultrasound images obtained by the minimum necessary operation of the ultrasound probe from information on the left parasternal border left ventricular long axis section obtained separately and the target image is extracted. Therefore, by using an ultrasound examination robot in combination, efficient operation of the ultrasound probe to automatically obtain the target image is possible, which significantly shortens the time required for echocardiography compared to the past and contributes to reducing the burden on the subject, and is also expected to solve the social issue of a shortage of echocardiography cases due to a shortage of examiners.

1 is a block diagram showing a schematic configuration of an echocardiography system according to an embodiment of the present invention. FIG. 2 is a conceptual diagram of an apical four-chamber cross-section for explaining a specific part of the heart. 13A to 13C are conceptual diagrams of an apical four-chamber cross section for explaining the relationship between the beam scanning plane and changes in the posture of an ultrasound probe in the yaw and roll directions in the desired posture estimation unit. 13A and 13B are conceptual diagrams of an apical four-chamber cross section for explaining the relationship between a posture change in the pitch direction of an ultrasound probe in a desired posture estimation unit and a beam scanning plane. 11 is a flowchart for explaining a procedure in a first step. 13 is a flowchart for explaining a procedure in a modified example of the first step. 10 is a flowchart for explaining a procedure in a second step. 13 is a flowchart for explaining a procedure in a third step.

The following describes an embodiment of the present invention with reference to the drawings.

1 shows a block diagram showing the schematic configuration of an echocardiography system including an ultrasound image search device according to this embodiment. In this figure, the echocardiography system 10 is a system that enables a doctor to perform an ultrasound examination (echocardiography) on a subject who is located away from the doctor. Specifically, the echocardiography system 10 includes an ultrasound image capture device 11 that captures ultrasound images depicting an examination cross section of the heart by contacting an ultrasound probe P with the subject's chest surface, a probe operation device 12 that operates the ultrasound probe P so that the position and posture relative to the subject can be changed, and a search device 13 that searches for an ultrasound image (hereinafter referred to as a "target image") depicting the apical four-chamber cross section, which is one of the basic cross sections for echocardiography, from multiple cardiac ultrasound images obtained while changing the position and posture of the ultrasound probe P in contact with the subject's chest surface.

A known ultrasound diagnostic device capable of acquiring ultrasound images for performing various ultrasound examinations is applied as the ultrasound imaging device 11. In other words, in this ultrasound imaging device 11, a two-dimensional tomographic image in the same cross section as the beam scanning plane F is acquired as an ultrasound image by beam scanning using the ultrasound probe P.

The probe operating device 12 is a robot that operates a holding part (not shown) that holds the ultrasonic probe P, thereby enabling six-degree-of-freedom motion control of the ultrasonic probe P, consisting of translational motion in three orthogonal axial directions and rotational motion around three orthogonal axes. Although not particularly limited, in this embodiment, an ultrasonic examination robot already proposed by the present inventors is applied (see JP 2023-70607 A). This ultrasonic examination robot is configured to be able to change the position and posture of the ultrasonic probe P placed relative to the chest of the subject while rotating the subject in a seated posture in the lateral and front-back directions. In addition, the ultrasonic examination robot is equipped with a mechanism for contacting the ultrasonic probe P with the chest surface with a predetermined pressing force in accordance with the unevenness of the chest surface of the subject. Note that the configuration of the ultrasonic examination robot is not an essential element of the present invention, so detailed description will be omitted. The above-mentioned operation control in the probe operating device 12 is performed by simple remote operation by an examiner such as a doctor, as well as by operation commands from a search device 13 required in the process of image search described later.

The ultrasonic probe P is provided with a known sensor (not shown) capable of detecting its position and orientation. This sensor is capable of acquiring the position of each coordinate of three orthogonal axes with a preset specific point as the origin, and the orientation, which is the rotation angle around the three axes. Although not particularly limited, the coordinate system of this embodiment is the probe coordinate system in the orientation of the ultrasonic probe P shown in FIG. 1. Note that as the sensor here, various devices and systems, such as an acceleration sensor, a magnetic sensor, an optical sensor, etc., can be used as long as they can acquire the position and orientation of the ultrasonic probe P.

The search device 13 is configured by a computer that includes a processor such as a CPU and storage devices such as a memory and a hard disk, and that functions as the following components. The search device 13 searches for a target image by estimating the appropriate position and posture of the ultrasound probe P that will obtain a target image depicting the desired apical four-chamber cross section.

This search device 13 includes a memory unit 15 that stores image data including ultrasound images acquired by the ultrasound imaging device 11, and an image analysis unit 16 that analyzes multiple ultrasound images of the heart acquired while changing the position and orientation of the ultrasound probe P to estimate the proper positioning of the ultrasound probe P for obtaining a target image, and extracts the ultrasound image obtained at the proper positioning as the target image.

The image data is data corresponding to the ultrasound image, including detection information acquired by image analysis described below, regarding the position and orientation of the ultrasound probe P when the ultrasound image was acquired from the detection results of the sensor attached to the ultrasound probe P, together with the acquisition time. The image data is stored sequentially in the memory unit 15 at each predetermined timing when the ultrasound probe P moves or rotates.

In the image analysis unit 16, the position and posture of the ultrasound probe P capable of acquiring an ultrasound image of the left parasternal border left ventricular long axis section among the examination sections used to diagnose a disease condition in echocardiography is set as the starting point of the search operation, and image analysis processing of the ultrasound image obtained with the subsequent operation of the ultrasound probe P is performed, and a target image is identified from among the multiple acquired ultrasound images. That is, here, the desired position that is estimated to be the subject's chest surface directly above the apex of the heart (hereinafter simply referred to as the "apex surface") is identified while translating the ultrasound probe P by the operation of the probe operation device 12. Then, the translation movement of the ultrasound probe P is stopped at the apex surface, and the desired posture at which the target image is obtained is identified while changing the posture of the ultrasound probe P by the operation of the probe operation device 12. These desired positions and desired postures are the appropriate placement of the ultrasound probe P at which the target image is obtained.

This image analysis unit 16 is composed of a starting point identification unit 18 that identifies the position and orientation of the ultrasound probe P at the starting point of the search operation, and a probe placement identification unit 19 that identifies the appropriate placement of the ultrasound probe P while changing the position and orientation of the ultrasound probe P from the starting point.

The origin identification unit 18 uses the method already proposed by the present inventors in JP 2023-70601 A to identify the position and orientation of the ultrasound probe P that will provide an optimal left parasternal border left ventricular long axis cross section, based on the ultrasound image acquired from the ultrasound probe P that is scanned along a predetermined range of the subject's chest surface by the operation of the probe operation device 12. In other words, according to this method, as shown in FIG. 2, the position and orientation of the ultrasound probe P that captures the mitral valve M in the cross section of the heart H at the center of the image is obtained, and this part is set as the origin S, and the position of the left ventricular long axis LL, which is the straight line connecting the apex A at the tip of the left ventricle H1 and the center of the mitral valve M, is obtained.

The probe placement identification unit 19 analyzes the ultrasound images obtained sequentially when the ultrasound probe P is translated and its posture is changed by the operation control of the probe operation device 12, thereby identifying the appropriate placement of the ultrasound probe P in the desired posture with the desired position on the surface of the subject's apex.

In analyzing this ultrasound image, detection information is obtained within the image of specific parts of the heart (mitral valve M, tricuspid valve T, ventricular septum VS) shown in FIG. 2. In other words, as the detection information here, an inference value indicating the probability of the presence (certainty) of the specific parts, their position within the image, etc. are appropriately obtained. For this image analysis, a known object detection model such as YOLO that utilizes deep learning to recognize specific objects based on trained image data is used. Note that for the image analysis here, various other systems and devices can be used as long as the detection information can be obtained based on the acquired ultrasound image.

The probe placement determination unit 19 includes a desired position estimation unit 21 that estimates the desired position of the ultrasound probe P that will be in contact with the apical surface, and a desired posture estimation unit 22 that estimates the desired posture of the ultrasound probe P that will obtain a target image at the desired position.

In the desired position estimation unit 21, the ultrasound probe P moves from the origin S along the left ventricular long axis L in the posture specified by the origin specification unit 18 while contacting the chest surface, according to the procedure described below. The vertical and horizontal (x-y axis) positions of the apical surface, which is the target position, are estimated by analyzing the ultrasound images acquired sequentially during the movement. In the image analysis here, the object detection model is used to obtain detection information such as the inferred value of the mitral valve M in each acquired ultrasound image, and the vertical and horizontal positions of the apical surface are identified based on the detection information. The position of the apical surface in the internal and external directions (z axis) is automatically identified because the probe operation device 12 has a mechanism for contacting the ultrasound probe P by following the unevenness of the subject. The ultrasound probe P located above the apical surface acquires an ultrasound image depicting the apical left ventricular long axis cross section.

The desired posture estimation unit 22 analyzes the ultrasound images acquired sequentially while changing the posture of the ultrasound probe P at the desired position on the apical surface of the heart, and estimates the desired posture of the ultrasound probe P that provides the best imaging conditions for the apical four-chamber cross-section: that is, the mitral valve M, ventricular septum VS, and tricuspid valve T are on the same plane, and that provides the best balance of the four chambers H1 to H4 of the heart H with the ventricular septum VS at the center.

This desired posture estimation unit 22 includes a yawing information acquisition unit 24 that identifies the desired posture in the yaw direction ψ by analyzing the ultrasound images obtained sequentially in accordance with the rotational movement in the yaw direction ψ of the ultrasound probe P abutting against the apex surface, a rolling information acquisition unit 25 that identifies the desired posture in the roll direction φ by analyzing the ultrasound images obtained sequentially in accordance with the rotational movement in the roll direction φ, and a pitching information acquisition unit 26 that identifies the desired posture in the pitch direction θ by analyzing the ultrasound images obtained sequentially in accordance with the rotational movement in the pitch direction θ.

The yawing information acquisition unit 24 analyzes the ultrasound images sequentially obtained when the ultrasound probe P is rotated in the yaw direction ψ from the initial position of the ultrasound probe P on the apex A shown in FIG. 3(A). In other words, here, the object detection model identifies the desired position of the ultrasound probe P in the yaw direction ψ, in which the beam scanning plane F contains more of the mitral valve M and ventricular septum VS, from each acquired ultrasound image, as shown in FIG. 3(B).

The rolling information acquisition unit 25 analyzes the ultrasound images obtained sequentially when the ultrasound probe P is rotated in the roll direction φ from the state shown in FIG. 3(B) in which the posture in the yaw direction ψ specified by the yawing information acquisition unit 24 is fixed. That is, for each acquired ultrasound image, as shown in FIG. 3(C), in order to depict more of the tricuspid valve T, which appears symmetrically with the mitral valve M centered on the ventricular septum VS, the desired posture in the roll direction φ of the ultrasound probe P is specified, in which the position of the ventricular septum VS becomes the beam scanning plane F that is depicted near the center of the image.

The pitching information acquisition unit 26 analyzes the ultrasound images obtained sequentially when the ultrasound probe P on the apex A is rotated in the pitch direction θ in the desired posture in the yaw direction ψ and roll direction φ specified by the yawing information acquisition unit 24 and the rolling information acquisition unit 25. That is, for each acquired ultrasound image, the desired posture in the pitch direction θ of the ultrasound probe P is specified, which is the beam scanning plane F that allows the four cavities H1 to H4 of the heart H to be depicted in the best balance, as shown in FIG. 4(B), from the state shown in FIG. 4(A).

In the above-described probe placement identification unit 19, the ultrasound image obtained by properly placing the ultrasound probe P is extracted as the target image.

Next, the procedure for searching for a target image while operating the ultrasound probe P will be explained using the flowcharts in Figures 5 to 8.

The following search procedure is executed in the following order: a first step is to identify the desired position of the ultrasonic probe P by image analysis processing in the desired position estimation unit 21; a second step is to identify the desired posture of the ultrasonic probe P in the yaw direction ψ and roll direction φ by image analysis processing in the yawing information acquisition unit 24 and the rolling information acquisition unit 25; and a third step is to identify the desired posture of the ultrasonic probe P in the pitch direction θ by image analysis processing in the pitching information acquisition unit 26.

In the first step, the desired location of the apical surface is identified using the procedure shown in Figure 5.

First, the origin S is identified by the origin identification unit 18 as the position of the ultrasound probe P in a posture that allows an appropriate left parasternal border left ventricular long axis cross section to be obtained (step S101). The ultrasound probe P moves in the same posture from the origin S to the left ventricular long axis LL (see FIG. 2) determined by the origin identification unit 18 while contacting the subject's chest surface (step S102) by the operation of the probe operation device 12. At this time, ultrasound images are acquired at predetermined time intervals (step S103). For each acquired ultrasound image, an inference value of the mitral valve M is calculated by the object detection model, and the presence or absence of detection of the mitral valve M is determined (step S104). Here, when the inference value is equal to or greater than zero or equal to or greater than a predetermined value, it is determined that the mitral valve M is present in the acquired ultrasound image. At this time, the position of the mitral valve M in the image is obtained, and it is determined whether or not the position in the image is near the center of the image, that is, within a predetermined allowable range (hereinafter referred to as the "image center range") from the center of the image (step S105). Therefore, when the position of the mitral valve M in the image does not exist within the image center range, the posture of the ultrasonic probe P is adjusted by rotating it in the roll direction φ by the operation control of the probe operation device 12 so as to eliminate the positional deviation depending on the distance in the left or right direction from the image center range in the image (step S106). The ultrasonic image acquired after this posture adjustment is stored in the storage unit 15 as image data corresponding to the position and posture of the ultrasonic probe P at that time (step S107). On the other hand, when the position of the mitral valve M in the image exists within the image center range, the ultrasonic image is stored in the storage unit 15 as image data corresponding to the position and posture of the ultrasonic probe P at that time (step S107). Then, it is determined whether the movement distance of the ultrasonic probe P from the starting point S has reached a preset distance (step S108). When the preset distance is reached, the position corresponding to the ultrasound image at the last timing when the inferred value of the mitral valve M is equal to or greater than a predetermined value, i.e., a preset threshold value, and the mitral valve M is detected multiple times consecutively, is identified as the desired position on the apical surface, and the ultrasound probe P is moved to the apical surface in this posture by the operation control of the probe operating device 12 (step S109). Here, if the mitral valve M is not detected multiple times consecutively, the position corresponding to the ultrasound image at which the inferred value of the mitral valve M is equal to or greater than the predetermined value and the mitral valve M was detected last is identified as the desired position. Note that the threshold here is not particularly limited, and may be an arbitrary preset value or the average value of the inferred values in each ultrasound image.

In addition, as the first step, the procedure shown in FIG. 6 can be used instead of the procedure shown in FIG. 5 to identify the desired position on the apical surface.

That is, here, the same processing as steps S101 to S107 in FIG. 5 described above is performed, but instead of the determination based on the moving distance of the ultrasound probe P in step S108, it is determined whether the non-detection state of the mitral valve M in step S104 continues a predetermined number of times (step S110). Then, when the non-detection state of the mitral valve M continues, the same processing as step S109 is performed.

In the first step described above, the process is based on the property that the entire ultrasound image suddenly turns black immediately after the ultrasound probe P, which moves along the left ventricular long axis LL with the mitral valve M as the starting point S, passes over the apex A due to the presence of an air layer between the ultrasound probe P and the heart H. In other words, based on the findings of the inventors' research that the moment just before the ultrasound image turns black is the position of the ultrasound probe P over the apex A, the position where the mitral valve M was last detected in the ultrasound image as the ultrasound probe P moves from above the mitral valve M along the left ventricular long axis LL is estimated to be the position of the apex surface.

In the second step shown in FIG. 7, the posture of the ultrasound probe P on the apical surface is appropriately changed, and the desired posture in the yaw direction ψ and roll direction φ of the ultrasound probe P is identified by processing in the yawing information acquisition unit 24 and the rolling information acquisition unit 25, in which the mitral valve M, tricuspid valve T, and ventricular septum VS are depicted on the same plane of the ultrasound image.

First, the ultrasound probe P moved to the apex surface in the first step is rotated in the yaw direction ψ by the operation of the probe operating device 12 (step S201), and an ultrasound image is acquired at a predetermined time interval (step S202). Then, the following process is performed on the acquired ultrasound image by the yawing information acquisition unit 24. That is, the object detection model calculates inference values including the presence or absence of detection of the ventricular septum VS and the mitral valve M as detection information (steps S203 and S204). Here, when the inference value is equal to or greater than zero or equal to or greater than a predetermined value, it is determined that the ventricular septum VS and the mitral valve M are present in the acquired ultrasound image. Then, an overall inference value is calculated by summing up the inference values for the ventricular septum VS and the mitral valve M, and the image data at that time including the detection information including the overall inference value is stored in the storage unit 15 (step S205). This overall inference value is calculated by summing up the values obtained by multiplying the inference values for the ventricular septum VS and the mitral valve M by a preset weighting constant. Then, when the rotation of the ultrasound probe P in the yaw direction ψ exceeds a predetermined angle (e.g., 180 degrees) (step S206), the rotation angle in the yaw direction ψ with the higher overall inference value is identified as the desired posture (step S207). Note that by comparing ultrasound images within a preset angle range (predetermined range), the rotation angle in the yaw direction ψ with the higher overall inference value of the ventricular septum VS and mitral valve M can be identified as the desired posture, and further, the highest rotation angle among the higher rotation angles within each range may be determined as the desired posture in the yaw direction ψ. Then, the probe operating device 12 rotates the ultrasound probe P so that it is in the desired posture in the yaw direction ψ.

Next, the ultrasonic probe P is rotated in the roll direction φ by the operation of the probe operation device 12 (step S208), and ultrasonic images are acquired at predetermined time intervals (step S209). Then, the following process is performed by the rolling information acquisition unit 24 for the acquired ultrasonic images. That is, the object detection model determines whether the ventricular septum VS is detected in the acquired ultrasonic images from its detection information (step S210). Here, too, when the inference value is equal to or greater than zero or equal to or greater than a predetermined value, it is determined that the ventricular septum VS is present in the acquired ultrasonic image, and it is determined whether the position of the ventricular septum VS is within the image center range (step S211). Therefore, if the position of the ventricular septum VS is not present within the image center range, the ultrasonic probe P is rotated in the roll direction φ by the operation control of the probe operation device 12 so that the ventricular septum VS is within the image center range (step S212). Then, in the process of rotating the ultrasound probe P in the roll direction φ while the orientation in the yaw direction ψ is fixed at the position of the apex surface, the ultrasound probe P at the rotation angle in the roll direction φ when the ventricular septum VS is initially within the image center range is identified as the desired orientation in the roll direction φ (step S213). The probe operating device 12 rotates the ultrasound probe P so that it assumes the desired orientation in the roll direction φ.

In the desired posture in the yaw direction ψ and roll direction φ identified in the second step described above, the mitral valve M is depicted with the ventricular septum VS in the center of the ultrasound image, and the tricuspid valve T, which is symmetrical to the mitral valve M across the ventricular septum VS, is also depicted on the same plane of the ultrasound image. Note that, as long as the position and posture of the ultrasound probe P at which there is always a moment when both the mitral valve M and the ventricular septum VS can be depicted when the ultrasound probe P is rotated fully around in the yaw direction ψ is known, the processing in the first step may be omitted, and the second step may be started from that position and posture.

In the third step shown in FIG. 8, the posture of the ultrasound probe P located on the apex surface is appropriately changed in the pitch direction θ while maintaining the desired posture in the yaw direction ψ and roll direction φ identified in the second step, and the desired posture of the ultrasound probe P in the pitch direction θ is identified when an image is obtained in which the four cavities H1 to H4 of the heart H are depicted in a better balanced state through processing by the pitching information acquisition unit 26.

First, the ultrasound probe P is rotated in the pitch direction θ by the operation of the probe operating device 12 while maintaining the posture in the yaw direction ψ and roll direction φ determined in the second step at the position of the apex surface (step S301), and ultrasound images are acquired at predetermined time intervals (step S302). The following processing is performed on the acquired ultrasound images by the pitching information acquisition unit 26. That is, the object detection model calculates inference values including the presence or absence of detection of the mitral valve M and tricuspid valve T as detection information (steps S303 and S304). Here, when the inference value is equal to or greater than zero or equal to or greater than a predetermined value, it is determined that the mitral valve M and tricuspid valve T are present in the acquired ultrasound image. Then, an overall inference value is calculated by summing up the inference values for the mitral valve M and tricuspid valve T, and the image data at that time including the detection information including the overall inference value is stored in the storage unit 15 (step S305). This overall inference value is calculated by summing values obtained by multiplying each inference value of the mitral valve M and the tricuspid valve T by a preset weighting constant. Then, it is determined whether the rotation of the ultrasonic probe P in the pitch direction θ is greater than or equal to a predetermined angle range (for example, a range of ±40 degrees based on the posture at the time of determining the desired position in the first step) (step S306), and when the rotation within the predetermined angle range is completed, the rotation angle in the pitch direction θ with the highest overall inference value is specified as the desired posture (step S307). Note that by comparing ultrasonic images within a preset angle range (predetermined range), the rotation angle in the pitch direction θ with the higher overall inference value of the mitral valve M and the tricuspid valve T can be specified as the desired posture. Furthermore, the highest rotation angle among the higher rotation angles within each range may be specified as the desired posture in the pitch direction θ.

The ultrasound image corresponding to the desired position and the desired posture of the ultrasound probe P identified through the above first to third steps is extracted as a target image, which is an ultrasound image depicting an apical four-chamber view suitable for examination (step S308).

In the above embodiment, the ultrasonic probe P is moved and rotated by the operation of the probe operation device 12, but the present invention is not limited to this. The operator can manually move and rotate the ultrasonic probe P under the guidance of the search device 13, and the desired position and desired posture of the ultrasonic probe P appropriate for obtaining a target image can be identified by processing in the search device 13.

In addition, the configuration and processing procedures of each part of the device in the present invention are not limited to the above description, and various modifications are possible as long as they produce substantially the same effect.

13 Search device 16 Image analysis unit 19 Probe placement identification unit 21 Desired position estimation unit 22 Desired posture estimation unit 24 Yawing information acquisition unit 25 Rolling information acquisition unit 26 Pitching information acquisition unit A Apex H1 to H4 Four chambers LL Left ventricular long axis M Mitral valve P Ultrasonic probe T Tricuspid valve VS Ventricular septum

Claims (7)

An apparatus for searching for a target image depicting a desired examination cross section for echocardiography from a plurality of cardiac ultrasound images obtained while changing the position and orientation of an ultrasound probe in contact with a chest surface of a subject, comprising:
an image analysis unit that estimates an appropriate arrangement of the ultrasound probe, which is a desired position and a desired posture, at which the target image is obtained, by image analysis of each of the ultrasound images, and extracts the ultrasound image obtained at the appropriate arrangement as the target image;
the image analysis unit performs image analysis for each of the ultrasound images to obtain detection information including an inference value representing a probability of the presence of a predetermined site in the heart that needs to be depicted in the target image, and includes a probe placement identification unit that identifies the proper placement of the ultrasound probe on the apex surface of the subject ;
the probe placement specifying unit includes a desired position estimating unit that estimates the desired position where the ultrasound probe comes into contact with the surface of the apex of the heart,
The desired position estimation unit, for each of the ultrasound images sequentially obtained while the ultrasound probe moves on the chest surface along the left ventricular long axis starting from a mitral valve whose position has been estimated in advance, determines, based on the detection information of the mitral valve, as the desired position the position of the ultrasound probe at the last timing when the inference value is equal to or greater than a predetermined value and the mitral valve is detected consecutively . An apparatus for searching for a target image depicting a desired examination cross section for echocardiography from a plurality of cardiac ultrasound images obtained while changing the position and orientation of an ultrasound probe in contact with a chest surface of a subject, comprising:
an image analysis unit that estimates an appropriate arrangement of the ultrasound probe, which is a desired position and a desired posture, at which the target image is obtained, by image analysis of each of the ultrasound images, and extracts the ultrasound image obtained at the appropriate arrangement as the target image;
the image analysis unit performs image analysis for each of the ultrasound images to obtain detection information including an inference value representing a probability of the presence of a predetermined site in the heart that needs to be depicted in the target image, and includes a probe placement identification unit that identifies the proper placement of the ultrasound probe on the apex surface of the subject;
the probe placement specifying unit includes a desired posture estimating unit that estimates the desired posture of the ultrasound probe that is in contact with the apex surface,
the desired posture estimation unit includes a yaw information acquisition unit that identifies the desired posture in a yaw direction by analyzing the ultrasound images sequentially obtained in accordance with a rotational movement of the ultrasound probe in the yaw direction,
The yawing information acquisition unit determines, for each of the ultrasound images, based on the detection information of the mitral valve and the ventricular septum, the desired posture, which is a posture in the yaw direction having a higher overall inference value, which is obtained by combining the inference values of the mitral valve and the ventricular septum. the desired attitude estimation unit includes a rolling information acquisition unit that, while fixing the desired attitude of the ultrasonic probe in the yaw direction specified by the yawing information acquisition unit, identifies the desired attitude in the roll direction by analyzing the ultrasonic images sequentially obtained in association with a rotational movement of the ultrasonic probe in the roll direction;
3. The ultrasound image searching device according to claim 2, wherein the rolling information acquisition unit determines, for each of the ultrasound images, a roll direction posture in which the ventricular septum is located near the center of the image as the desired posture based on the detection information of the ventricular septum . The desired attitude estimation unit includes a pitching information acquisition unit that specifies the desired attitude in the pitch direction by analyzing the ultrasonic images sequentially obtained in association with a rotational movement of the ultrasonic probe in the pitch direction while fixing the desired attitude in the yaw direction and the roll direction of the ultrasonic probe specified by the yawing information acquisition unit and the rolling information acquisition unit,
4. The ultrasonic image searching device according to claim 3, wherein the pitching information acquisition unit determines, for each of the ultrasonic images, a posture in the pitch direction having a higher overall inference value obtained by combining the inference values of the mitral valve and the tricuspid valve based on the detection information of the mitral valve and the tricuspid valve as the desired posture. An apparatus for searching for a target image depicting a desired examination cross section for echocardiography from a plurality of cardiac ultrasound images obtained while changing the position and orientation of an ultrasound probe in contact with a chest surface of a subject, comprising:
an image analysis unit that estimates an appropriate arrangement of the ultrasound probe, which is a desired position and a desired posture, at which the target image is obtained, by image analysis of each of the ultrasound images, and extracts the ultrasound image obtained at the appropriate arrangement as the target image;
the image analysis unit performs image analysis for each of the ultrasound images to obtain detection information including an inference value representing a probability of the presence of a predetermined site in the heart that needs to be depicted in the target image, and includes a probe placement identification unit that identifies the proper placement of the ultrasound probe on the apex surface of the subject;
the probe placement specifying unit includes a desired posture estimating unit that estimates the desired posture of the ultrasound probe that is in contact with a surface of the apex of the heart so that an apical four-chamber cross section is depicted as the target image;
The desired posture estimation unit specifies the desired posture in the yaw direction and the roll direction, which enables the mitral valve, the tricuspid valve, and the ventricular septum to be visualized on the same plane of the ultrasound image and enables the four chambers to be visualized with the ventricular septum at the center, based on the detection information of the mitral valve and the ventricular septum obtained by analysis of each of the ultrasound images obtained sequentially in accordance with the rotational movement of the ultrasound probe in the yaw direction and the roll direction, and specifies the desired posture in the pitch direction, which enables the four chambers to be visualized in a more balanced state, from the detection information of the mitral valve and the tricuspid valve obtained by analysis of each of the ultrasound images obtained sequentially in accordance with the rotational movement of the ultrasound probe in the pitch direction in the desired posture state in the yaw direction and the roll direction. A program for an apparatus for searching for a target image depicting a desired examination cross section for an echocardiogram from a plurality of cardiac ultrasound images obtained while changing the position and orientation of an ultrasound probe in contact with a chest surface of a subject, the program comprising:
by performing image analysis of each of the ultrasound images, a proper arrangement of the ultrasound probe, which is a desired position and a desired posture, at which the target image is obtained, is estimated, and the ultrasound image obtained at the proper arrangement is extracted as the target image;
the image analysis unit performs image analysis for each of the ultrasound images to obtain detection information including an inference value representing a probability of the presence of a predetermined site in the heart that needs to be depicted in the target image, and includes a probe placement identification unit that identifies the proper placement of the ultrasound probe on the apex surface of the subject ;
the probe placement specifying unit includes a desired position estimating unit that estimates the desired position where the ultrasound probe comes into contact with the surface of the apex of the heart,
The desired position estimation unit, for each of the ultrasound images sequentially obtained while the ultrasound probe moves on the chest surface along the left ventricular long axis starting from a mitral valve whose position has been estimated in advance, determines, based on the detection information of the mitral valve, as the desired position, the position of the ultrasound probe at the last timing when the inference value is equal to or greater than a predetermined value and the mitral valve is detected consecutively . A program for an apparatus for searching for a target image depicting a desired examination cross section for an echocardiogram from a plurality of cardiac ultrasound images obtained while changing the position and orientation of an ultrasound probe in contact with a chest surface of a subject, the program comprising:
by performing image analysis of each of the ultrasound images, a proper arrangement of the ultrasound probe, which is a desired position and a desired posture, at which the target image is obtained, is estimated, and the ultrasound image obtained at the proper arrangement is extracted as the target image;
the image analysis unit performs image analysis for each of the ultrasound images to obtain detection information including an inference value representing a probability of the presence of a predetermined site in the heart that needs to be depicted in the target image, and includes a probe placement identification unit that identifies the proper placement of the ultrasound probe on the apex surface of the subject;
the probe placement specifying unit includes a desired posture estimating unit that estimates the desired posture of the ultrasound probe that is in contact with a surface of the apex of the heart so that an apical four-chamber cross section is depicted as the target image;
The desired posture estimation unit specifies the desired posture in the yaw direction and the roll direction, which enables the mitral valve, tricuspid valve, and ventricular septum to be visualized on the same plane of the ultrasound image and enables the four chambers to be visualized with the ventricular septum at the center, based on the detection information of the mitral valve and the ventricular septum obtained by analysis of each of the ultrasound images obtained sequentially in accordance with the rotational movement of the ultrasound probe in the yaw direction and the roll direction, and specifies the desired posture in the pitch direction, which enables the four chambers to be visualized in a more balanced state, from the detection information of the mitral valve and the tricuspid valve obtained by analysis of each of the ultrasound images obtained sequentially in accordance with the rotational movement of the ultrasound probe in the pitch direction in the desired posture state in the yaw direction and the roll direction.

Images