JP5526401B2 - Ventricular wall information extraction device - Google Patents

Description

  The present invention relates to a ventricular wall information extracting device that extracts information such as the outline, shape, and three-dimensional position of the ventricular wall of the heart.

  The ultrasonic diagnostic method is an indispensable technique for diagnosing the heart because of its safety and real-time property because an ultrasonic tomographic image of the patient's body can be obtained simply by pressing the probe onto the patient's body surface. The progress of miniaturization and portability of the ultrasonic diagnostic apparatus used in this ultrasonic diagnostic method is remarkable, and ultrasonic diagnostics are realized anywhere.

  On the other hand, by increasing the network transmission speed such as an inter-nut, it is possible to transmit an ultrasonic tomographic image captured by a probe of an ultrasonic diagnostic apparatus via a network. Under such a technical background, research on a remote ultrasonic diagnostic method in which a hospital doctor performs an ultrasonic diagnosis on a remote patient based on an ultrasonic tomographic image transmitted from a patient's home to the hospital is being promoted. .

  In this type of remote ultrasound diagnosis, a remote operation using a probe operation robot is used to accurately transmit the position and posture of a probe desired by a doctor to a remote place to obtain an accurate ultrasonic tomographic image of a patient. Ultrasound diagnostic methods have been reported. Also, using AR (Augmented Reality) technology called Augmented Reality, CG (computer graphic) representing the position and posture of the probe is projected onto the patient's body surface, and the probe operation is performed on the patient's operator. Research that teaches is underway.

  On the other hand, regarding an ultrasonic diagnostic apparatus, an ultrasonic diagnostic apparatus is proposed that can extract the contour of an object in an ultrasonic tomographic image with high accuracy without forcing the operator to perform many operations. (For example, refer to Patent Document 1). This ultrasonic diagnostic apparatus includes an initial contour extraction unit that roughly extracts a contour of an object in a tomographic image as an initial contour by performing image processing such as equalization, binarization, and degeneration on the tomographic image; By using a dynamic contour model such as SNAKES applied to an object in a tomographic image with the initial contour as an initial value, an automatic contour extraction unit including a dynamic contour extraction unit that finely extracts a contour is provided. Is.

  As for the image processing device, an image processing device that enables accurate and efficient contour extraction as a result of easily setting an accurate initial contour when extracting a dynamic contour of an object in an image is proposed. (See, for example, Patent Document 2). The image processing apparatus is configured to generate an initial contour of an object based on a feature point of the object in the image with respect to the input image, and to change the contour shape of the object without changing the position of the feature point. Contour extraction that extracts the contour of the object in the image by modifying the initial contour so that the sum of the feature amount, the image feature amount, and the initial contour feature amount given as necessary is minimized. Means.

JP 2002-224116 A (summary) JP 11-265455 A (summary)

  By the way, in the remote ultrasonic diagnostic method using a robot for probe operation, a certain result is expected in the field of emergency medicine where a robot can be installed, but the patient's home where it is difficult to install the robot The feasibility in etc. is thin. Further, in a remote ultrasonic diagnostic method in which a CG representing the position / posture of the probe is projected onto the patient's body surface and the patient's operator is instructed to operate the probe, the doctor's instructions must be sequentially reflected via the network. Therefore, it takes time to obtain a target ultrasonic tomographic image as compared with a normal ultrasonic diagnosis, and it is difficult for a doctor to grasp a three-dimensional space.

  On the other hand, in the ultrasonic diagnostic apparatus described in Patent Document 1, a dynamic contour extraction method such as SNAKES is applied. However, many of the conventional contour extraction methods using SNAKES generally have initial contours. Is placed in the vicinity of the contour of the target object to achieve convergence, the user needs to place the initial contour for each image, and it takes time to search for the initial contour. There is a problem. Similarly, in the image processing apparatus described in Patent Document 2, it is necessary to manually set the initial contour, which places a heavy burden on the operator.

  That is, in order to extract the outline of the ventricular wall of the heart using a conventional ultrasonic diagnostic apparatus and image processing apparatus, the probe operator is set with the whole or part of the region where the ventricular wall is drawn as the initial position. However, it is necessary to specify manually, which places a heavy burden on the operator of the probe.

The present invention has been made in response to such problems of the prior art, it can be initialized, such as the initial contour at all unnecessary and Oh while seeking to easily extract the ventricular wall information of the heart, As a result, an object of the present invention is to provide a ventricular wall information extraction device that can easily realize remote ultrasonic diagnosis.

In order to solve such problems, a ventricular wall information extraction device according to the present invention is a device that extracts ventricular wall information based on an ultrasonic tomographic image of the heart imaged by a probe of an ultrasonic diagnostic apparatus. And dividing the ultrasonic tomographic image into a plurality of small regions, defining an average vector in each small region as a region motion vector f of each small region, and calculating the motion vector by the following equation (10), the ventricular wall A motion vector measuring means for measuring a plurality of motion vectors,

(V _n is the calculated vector, N is the number of vectors in the small area)
A motion vector measuring means for measuring a plurality of motion vectors of the ventricular wall, an initial search point setting means for setting an intersection region of extension lines of the plurality of motion vectors measured by the motion vector measuring means as an initial search point, and an initial search point It further comprises contour point extracting means for extracting a plurality of contour points of the ventricular wall by polar coordinates with the initial search point set by the setting means as the origin.

  In the ventricular wall information extracting device according to the present invention, when an ultrasonic tomographic image of the ventricular wall of the heart is captured by the probe of the ultrasonic diagnostic apparatus, the motion vector measuring means measures a plurality of motion vectors of the ventricular wall and performs an initial search. The point setting means sets an intersection area of extension lines of a plurality of motion vectors as an initial search point. Then, the contour point extracting means extracts a plurality of contour points of the ventricular wall based on the polar coordinates having the set initial search point as the origin. That is, the initial search point setting means automatically sets the initial search point, even though the probe operator does not initialize the initial contour of the ventricular wall, and the contour point extracting means extracts a plurality of contour points of the ventricular wall. To do.

  In the ventricular wall information extracting device of the present invention, the ellipse approximating means for drawing an approximate ellipse along the ventricular wall contour from a plurality of contour points of the ventricular wall extracted by the contour point extracting means by the least square method, It is possible to add a three-dimensional measuring means for obtaining information on the shape and three-dimensional position of the ventricle wall based on a plurality of approximate ellipses drawn according to the random operation of the probe.

  In this invention, when the contour point extracting means extracts a plurality of contour points of the ventricular wall, the ellipse approximating means draws an approximate ellipse along the contour of the ventricular wall from the plurality of contour points, and the plurality of contour points are extracted according to the random operation of the probe. When the approximate ellipse is drawn, the three-dimensional measuring means obtains information on the shape and three-dimensional position of the ventricular wall based on the plurality of approximate ellipses. Here, since the approximate ellipse drawn by the ellipse approximation means is based on the least square method with a small calculation processing load, the approximate ellipse is drawn in real time.

  According to the ventricular wall information extracting device according to the present invention, when an ultrasonic tomographic image of the ventricular wall of the heart is captured by the probe of the ultrasonic diagnostic apparatus, the motion vector measuring means measures a plurality of motion vectors of the ventricular wall, Since the initial search point setting means sets the intersection region of the extension lines of a plurality of motion vectors as the initial search point, the initial setting of the initial contour of the ventricular wall, which has conventionally been required to be manually set by the probe operator Etc. are completely unnecessary. Then, since the contour point extracting means extracts a plurality of contour points of the ventricular wall from the polar coordinates with the initial search point set by the initial search point setting means as the origin, ventricular wall information including the contour of the ventricular wall is easily extracted. can do. That is, according to the present invention, it is possible to easily extract ventricular wall information including the contour of the ventricular wall, while no initial setting such as initial contour is required, and as a result, remote ultrasonic diagnosis is also easily realized. It becomes possible to do.

  Further, according to the ventricular wall information extracting device of the present invention to which the ellipse approximating means and the three-dimensional measuring means are added, when the contour point extracting means extracts a plurality of contour points of the ventricular wall, the ventricular wall is extracted from the plurality of contour points. When the ellipse approximating means draws an approximate ellipse along the contour, and when a plurality of approximate ellipses are drawn in response to the random operation of the probe, the three-dimensional measuring means based on the plurality of approximate ellipses Since the information on the three-dimensional position is obtained, the ventricular wall information including the shape of the ventricular wall and the three-dimensional position can be easily extracted, and the remote ultrasonic diagnosis can be easily realized.

It is a functional block diagram which shows the structure of the ventricle wall information extraction apparatus which concerns on one Embodiment of this invention. It is a front view of the heart which shows the position and direction of the probe which image the long-axis cross section of the left ventricle made into the object by the ventricle wall information extraction apparatus which concerns on one Embodiment. FIG. 3 is a long-axis cross-sectional view of the left ventricle imaged by the probe shown in FIG. 2. It is a front view of the heart which shows the position and direction of the probe which images the short-axis cross section of a left ventricle. FIG. 5 is a short-axis cross-sectional view of the left ventricle imaged by the probe shown in FIG. 4. It is the model which assumed the left ventricle to be an ellipsoid. It is a figure which shows the ultrasonic tomographic image of the short axis direction crossing a left ventricle. FIG. 8 is a diagram in which the ultrasonic tomographic image shown in FIG. 7 is divided into small regions and region motion vectors are superimposed and displayed. It is a figure which shows the polar coordinate which extracts the outline point of the left ventricle wall. It is a figure which shows the approximate ellipse along the outline point of the left ventricle wall. It is a two-dimensional projection figure of a general ellipse. It is a graph which shows the distance error of the ellipse center point calculated from N sample points, and a true center point. FIG. 8 is a diagram in which an ellipse approximated by an ellipse approximation unit is superimposed and displayed on the ultrasonic tomographic image in the short axis direction of the left ventricle shown in FIG. 7. It is a figure which shows the ultrasonic tomographic image by which the mitral valve and the right ventricle were drawn. FIG. 15 is a diagram in which an elliptical shape approximated based on the ultrasonic tomographic image illustrated in FIG. 14 is superimposed and displayed on the ultrasonic tomographic image. And d _n value and the variance value d _sigma shown in FIG. 15 is a graph showing the relationship between the axial rotation angle of the probe. FIG. 8 is a diagram in which a region of the left ventricle wall is superimposed and displayed on an ultrasonic tomographic image in the short axis direction of the left ventricle shown in FIG. 7. FIG. 8 is a diagram in which a motion vector extracted from the left ventricular wall is superimposed and displayed on an ultrasonic tomographic image of the left ventricle in the short axis direction shown in FIG. 7. It is the figure which superimposed and displayed the ellipse approximated using the Biweight method and the ellipse approximated without using the Biweight method on the ultrasonic tomographic image of the short axis direction of the left ventricle. It is a perspective view which shows the arrangement | positioning relationship between a patient, a camera, the marker of a probe, and a monitor. It is a figure which shows the GUI screen projected on the monitor of FIG. It is a figure which shows the other GUI screen projected on the monitor of FIG. It is a figure which shows the positional relationship of an arbitrary ellipse center point and an ultrasonic tomographic plane. It is a figure which shows the estimation and the presentation state of the long-axis cross section of the left ventricle wall for one test subject. It is a graph which shows the result of having approximated to the straight line by the least square method from the distribution of the ellipse center point obtained about each 8 subjects, and calculating the average distance and the standard deviation from the straight line to each ellipse center point.

  Hereinafter, embodiments of a ventricular wall information extracting device according to the present invention will be sequentially described with reference to the accompanying drawings. As shown in FIG. 1, a ventricular wall information extracting device 1 according to an embodiment is a device that extracts ventricular wall information based on an ultrasonic tomographic image of the heart imaged by a probe 2 </ b> A of an ultrasonic diagnostic device 2. For example, information such as the outline, shape, and three-dimensional position of the left ventricle that discharges fresh blood flowing from the pulmonary veins via the left atrium to the aorta is extracted.

  The ultrasonic diagnostic apparatus 2 includes a probe 2A that is operated by being pressed against the patient's body surface, an image generation unit 2B that generates an ultrasonic tomographic image in the patient's body based on a signal searched by the probe 2A, and an image And a monitor 2C for displaying the ultrasonic tomographic image generated by the generation unit 2B.

  Here, the ventricular wall information extracting device 1 of one embodiment cooperating with the ultrasonic diagnostic apparatus 2 includes a motion vector calculating unit (motion vector measuring means) 1A that measures at least a plurality of motion vectors of the left ventricular wall, and a motion vector. An initial search point setting unit (initial search point setting means) 1B that sets an intersection region of extension lines of a plurality of motion vectors measured by the calculation unit 1A as an initial search point, and an initial set by the initial search point setting unit 1B A contour point extraction unit (contour point extraction means) 1C that extracts a plurality of contour points of the left ventricular wall by polar coordinates with the search point as the origin is provided.

  Further, the ventricular wall information extracting apparatus 1 includes an ellipse approximating unit for drawing an approximate ellipse along the contour of the left ventricular wall from a plurality of contour points of the ventricular wall extracted by the contour point extracting unit 1C by the least square method ( Ellipse approximating means) 1D and a three-dimensional measuring unit (3) for obtaining information on the shape and three-dimensional position of the left ventricular wall based on a plurality of approximate ellipses drawn according to the random operation of the probe 2A of the ultrasonic diagnostic apparatus 2 Dimension measuring means) 1E.

The motion vector calculation unit 1A, the initial search point setting unit 1B, the contour point extraction unit 1C, the ellipse approximation unit 1D, and the three-dimensional measurement unit 1E of the ventricular wall information extraction device 1 are an input / output interface I / O, an A / D converter, ROM (Read Only Memory) that stores programs and data, RAM (Random Access Memory) that temporarily stores input data, etc., hardware such as a personal computer (not shown) that includes a CPU (Central Processing Unit) that executes programs, and the like; It is configured using software and cooperates with the ultrasonic diagnostic apparatus 2.
<Long-axis and short-axis cross sections of the left ventricle>

  Here, there are a long-axis cross-section, a short-axis cross-section, a four-chamber cross-section, and the like as cross-sections that serve as a basis for diagnosis of the heart. The long-axis cross-section is a cross-section obtained by imaging the left ventricle (Left ventricle) in the long-axis direction with the probe 2A operated in the position and orientation as shown in FIG. 2, and as shown in FIG. The posterior wall PW and the interventricular septum IVS are viewed in parallel, so the left ventricular diameter is maximized. At the same time, the mitral valve MV and the aortic valve AV are observed. .

  On the other hand, as shown in FIG. 4, the short-axis cross-section is a cross-section viewed by minimizing the cross-sectional area of the left ventricular wall LV by rotating the probe 2A shown in FIG. 2 about 90 degrees clockwise. As shown in FIG. 5, in the case of a normal heart, the left ventricular wall LV is imaged concentrically, and the mitral valve MV opens symmetrically.

  When the contact position of the probe 2A when such a short-axis cross-section is acquired and the probe 2A is rotated around the axis, the cross-sectional shape of the left ventricular wall LV is usually deformed from a circle to an ellipse. .

  Here, assuming that the three-dimensional shape of the left ventricle LV is an ellipsoid, as shown in FIG. 6, the cross-sectional shape of the wall of the left ventricle LV becomes an ellipse in an arbitrary cross section crossing the left ventricle LV. Therefore, the position of the probe 2A for obtaining the longest cross section, that is, the long-axis cross section, can be estimated by measuring the spatial position of a plurality of cross sections.

That is, the three-dimensional position of the probe 2A is constantly monitored, and when the cross section crossing the left ventricle LV is recognized as an ellipse, the contour of the wall of the left ventricle LV is extracted in real time by image processing, and the ellipse E _i To approximate. At this time, the three-dimensional coordinates P _i of the center point of the ellipse are calculated, and similarly, the center point of the ellipse obtained from the positions of the plurality of probes 2A is acquired. Since the cross-section from which the long-axis cross section is obtained is considered to be close to these points, the range of spaces that are candidates for the long-axis cross-section can be limited.
<Left ventricular wall contour extraction>

  FIG. 7 shows an ultrasonic tomographic image in the short axis direction crossing the left ventricle LV. It is difficult to extract the wall of the left ventricle LV only by the binarization process using a simple threshold because the attenuation characteristic of the ultrasonic wave in the depth direction from the body surface is not uniform. Therefore, the ventricular wall information extracting device 1 of the embodiment shown in FIG. 1 extracts the contour of the wall of the left ventricle LV in cooperation with the ultrasonic diagnostic device 2.

First, as shown in FIG. 8, an ultrasonic tomographic image is divided into a plurality of small areas of 15 × 15 [pixel], for example, and an average vector in each small area is defined as a region motion vector f of each small area. The motion vector calculation unit 1A calculates the following formula 1. Here, V _n is a calculated vector, and N is the number of vectors in the small area.

Subsequently, the region motion vector f calculated by the motion vector calculation unit 1A is extended to the end point infinitely during the systole of the left ventricle LV, and the start point is extended infinitely during the diastole of the left ventricle LV. The point setting unit 1B sets those intersection areas as initial search points O (see FIG. 9). At this time, since the calculated region of intersection is the region motion vector f calculated only in the direction, the initial search point can be obtained by adding the absolute value (size) of the region motion vector f as a weight. The setting unit 1B sets an appropriate initial search point O. That is, the initial search point setting unit 1B defines an index q obtained by adding a weight to the number of intersections p in the small region by the following formula 2.

  On the other hand, since the number of vectors calculated by the gain of the ultrasonic tomographic image fluctuates and affects the q value itself, the total number N of intersections is normalized to N = A (A-1) / 2, and the intersection index I Is defined as I = q / N = 2q / A (A-1).

  Next, as shown in FIG. 9, the contour point extraction unit 1C uses the polar coordinates having the radius r with the initial search point O set by the initial search point setting unit 1B as the origin, and the contour of the wall of the left ventricle LV. Extract multiple points. That is, by changing the radial angle θ of the radius r by 1 [deg], the maximum value of luminance is extracted from the luminance distribution on 360 radials, and the wall of the left ventricle LV is determined by the maximum value distribution on the radials. Extract contour points.

At that time, the contour point extracting unit 1C extracts a specific region by adding a morphological operation as an image processing method for extracting a shape feature and adding a label to each pixel as an attribute. By applying a labeling process for classifying a plurality of regions as a group by adding the same label, a contour point of the wall of the left ventricle LV from which the minute area region is removed is extracted.
<Ellipse approximation of left ventricular wall outline>

Subsequently, as shown in FIG. 10, the ellipse approximation unit 1D approximates the set of contour points of the wall of the left ventricle LV extracted by the contour point extraction unit 1C to an ellipse using the least square method. Here, a general two-dimensional projection view of an ellipse is shown in FIG. Where (X _n , Y _n ) are the coordinates of the nth contour point, (X ₀ , Y ₀ ) are the center coordinates of the ellipse, a and b are the major and minor radii of the ellipse, and φ is the inclination of the ellipse. ing.

  In general, Hough transform, least square method, etc. can be considered as a method for determining a figure parameter from a set of a plurality of points. However, in Hough transform, the calculation processing load on the number of parameters is large, and real-time processing is difficult. In the embodiment, the least square method is introduced with emphasis on calculation cost.

Here, Formula 3 is developed, and the sum of squares at all the contour points is calculated as Formula 4. Here, all of A to E are coefficients represented by elliptic parameters X ₀ , Y ₀ , a, b, and φ, and N is the number of contour points.

Formula 5 is obtained by partial differentiation of this with respect to each of A to E and formulating it to be zero. Then, by applying the inverse matrix of the first term on the left side of Formula 5 to both sides from the left, A to E are obtained, so that each parameter X ₀ , Y ₀ , a, b, φ can be obtained by inverse calculation.

<Reliability of elliptical approximation>

In order to confirm the reliability of the above ellipse approximation, first, the parameters of the ellipse are set in advance as X ₀ = 50, Y ₀ = 30, a = 15, b = 10 [mm], and φ = 10 [deg]. N sample points are taken at equal intervals from the orbit of the ellipse.

  Next, a distance error between the ellipse center point calculated from the N sample points and the true center point is obtained. The distance error is obtained when the value of N is changed, and the result of the transition is shown in FIG.

From this result, it was confirmed that the center point can be obtained with an error within 1 [mm] if there are 8 or more points at equal intervals on the orbit of the ellipse. Here, since the extracted contour points of the wall of the left ventricle LV are 300 or more as shown in FIG. 10, it is considered that the necessary number of data is sufficiently satisfied. FIG. 13 shows an ellipse approximated by the ellipse approximation unit 1D superimposed on the ultrasonic tomographic image in the short axis direction of the left ventricle LV shown in FIG. 7, and the ellipse follows the shape of the wall of the left ventricle LV. It is drawn accurately.
<Correction of misrecognition by the Biweight method>

  By the way, in the above process, when structures other than the wall of the left ventricle LV are included, an ellipse greatly deviating from the contour of the wall of the left ventricle LV is drawn. For example, when an ultrasonic tomographic image in which the mitral valve MV and the right ventricle RV (Right ventricle) are clearly depicted as shown in FIG. 14 is processed, an approximate elliptical shape f (x, y) is shown in FIG. As such, it will deviate greatly from the actual.

  Therefore, in order to exclude shapes other than the wall of the left ventricle LV such as the mitral valve MV and the right ventricle RV, the following two evaluation criteria are provided for the results obtained by the ellipse approximation to discriminate them. did.

  First, the ratio of the major radius a and the minor radius b shown in FIG. 11 was derived for the extracted ellipse. This is because when an outline point other than the wall of the left ventricle LV occurs, the drawn ellipse is distorted in either direction from the shape of the wall of the left ventricle LV, and the effect appears in a and b.

Here, since the number of center points necessary for estimation of the long-axis cross section is approximately 8, the ellipse center points are sampled from the top eight cross sections with a / b close to 1 among the cross sections on which the ellipse is drawn. Decided to do. Secondly, the degree of coincidence d _ave and d _σ between each extracted contour point and the ellipse approximated _therefrom was evaluated.

Here, contour extraction points shown in FIG. 15 (X _n, Y _n) and the ellipse center point (X _0, Y ₀₎ and the straight line y = a _n x + b _n connecting the, drawn ellipse f (x, y the distance between the intersection of) the d _n, the average of d _n calculated for all sampled points was d _ave. Further, the d _sigma represents the variance of d _n.

These values were verified using actual ultrasonic tomographic images. Targeting healthy males in their 20s, an ultrasonic tomographic image of the heart is acquired in increments of 15 [deg] from -90 to 90 [deg] with the axis rotation angle of the probe 2A capable of imaging a short-axis cross section being 0 [deg]. In each case, d _ave and d _σ were measured. As a result, as shown in FIG. 16, both d _ave and d _σ were within 2 [mm] in the range of ± 30 [deg]. Therefore, an ultrasonic tomographic image satisfying d _ave and d _σ ≧ 2 [mm] is excluded, and the elliptical center point is restricted from being stored.

  Here, in the ventricular wall information extraction device according to the embodiment, the three-dimensional measurement unit 1E is configured to detect the left ventricle LV based on a plurality of approximate ellipses drawn according to the random operation of the probe 2A of the ultrasonic diagnostic device 2. Measure wall shape and 3D position.

  Also, as shown in FIG. 17, since only the wall region of the left ventricle LV can be automatically recognized, the center point from the elliptical shape is corrected and applied to the contour recognition of the next frame, as shown in FIG. In addition, it is possible to extract a motion vector only from the wall of the left ventricle LV in a short time. As a result, the video signal can be displayed in real time, and automatic evaluation of the heart's motor function is also possible.

  As described above, in the ventricular wall information extracting apparatus 1 according to the embodiment, when an ultrasonic tomographic image of the wall of the left ventricle LV of the heart is captured by the probe 2A of the ultrasonic diagnostic apparatus 2, the motion vector calculating unit 1A Measures a plurality of motion vectors of the wall of the left ventricle LV, and the initial search point setting unit 1B sets the intersection region of the extension lines of the plurality of motion vectors as the initial search point O.

  Then, the contour point extraction unit 1C extracts a plurality of contour points of the left ventricle LV with polar coordinates having a radius r of the radius r with the set initial search point O as the origin. That is, the initial search point setting unit 1B automatically sets the initial search point O and the contour point extraction unit 1C sets the left ventricle even though the operator of the probe 2A does not perform the initial setting of the initial contour of the left ventricle LV. A plurality of contour points of the LV wall are extracted.

  Therefore, according to the ventricular wall information extracting apparatus 1 of the embodiment, there is no need for the initial setting of the initial contour of the wall of the left ventricle LV, which has conventionally been required to be manually set by the operator of the probe 2A. Become. Then, the contour point extraction unit 1C extracts a plurality of contour points of the wall of the left ventricle LV from the polar coordinates having the radius r of the initial search point O set by the initial search point setting unit 1B. Ventricular wall information including the outline of the wall of the left ventricle LV can be easily extracted. That is, it is possible to easily extract ventricular wall information including the contour of the wall of the left ventricle LV without any initial setting such as initial contour, and as a result, remote ultrasonic diagnosis can be easily realized. It becomes possible.

  Further, in the ventricular wall information extracting device 1 according to the embodiment, when the contour point extracting unit 1C extracts a plurality of contour points of the wall of the left ventricle LV, approximation is performed from the plurality of contour points along the contour of the wall of the left ventricle LV. When the ellipse is drawn by the ellipse approximator 1D and a plurality of approximate ellipses are drawn according to the random operation of the probe 2A, the ellipse approximator 1D draws the shape of the wall of the left ventricle LV and 3 Measure the dimension position.

  Therefore, according to the ventricular wall information extracting device 1 of one embodiment, ventricular wall information including the shape and three-dimensional position of the wall of the left ventricle LV can be easily extracted, and remote ultrasonic diagnosis is also easily realized. It becomes possible.

  The ventricular wall information extraction device of the present invention is not limited to the above-described embodiment. For example, for an ellipse that does not fall within the aforementioned threshold range, a more stable ellipse can be derived using the Biweight method.

Each extraction point and by weight (Biweight) method (X _n, Y _n) respectively in accordance with the distance d _n between the drawn ellipse as a technique for imparting a weight w (d _n), a more robust approximation I can expect. Equations 6 and 7 show how to calculate the weight w (d _n ). However, W is an allowable range of error, and here it is 4 [mm], which is the sum of the threshold _values of d _ave and d _σ described above.

FIG. 19 shows the result obtained by assigning the weight w (d _n ) calculated by Equation 6 and Equation 7 to each contour point, performing elliptical approximation again by the least square method, and displaying the result superimposed on the ultrasonic tomographic image. It can be confirmed that an ellipse close to the shape of the left ventricle LV is drawn as compared with the case where the Biweight method is not used.
<Acquisition of long-axis cross section using AR>

  In order to convert the distribution of the ellipse center point extracted by the above method into the three-dimensional space, it is necessary to grasp the absolute three-dimensional position / posture of the probe 2A. However, magnetic and optical three-dimensional position sensors are generally expensive and are not suitable for use at home due to the complexity of the installation method.

  Therefore, in one embodiment, ARToolKit is used as a simple three-dimensional position measurement method. This is a library that can measure the position and orientation in real time using only the camera and markers printed on paper, and the main cost is only the camera. The principle of operation is to detect a plurality of markers with a pattern drawn inside a square by applying image processing such as binarization and labeling to an image shot with a fixed USB camera, and then the internal pattern. The relative position and angle are calculated from the size and orientation of the.

  Therefore, in one embodiment, as shown in FIG. 20, it is assumed that the operator of the probe 2A sits on the right hand side of the patient and operates with the right hand, and the markers are positioned above the probe 2A and the center of both shoulder clavicles of the patient. I decided to install it in the vicinity. The camera was placed above the right shoulder of the patient so that the entire chest could be photographed and the marker was not easily hidden by the hand or probe. Thereby, the probe tip position with the patient body surface as the origin can be calculated in real time from the relative relationship between the two with respect to the camera coordinate system, without depending on the installation position of the camera.

  FIG. 21 and FIG. 22 show GUI (Graphical User Interface) screens displayed on the monitor 2C of FIG. Visual C ++ was used to develop the program, DirectShow was used to acquire the video images, and OpenGL was used to display the graphics. An ultrasonic tomographic image and a patient video are displayed in the window, and the operator can observe the hand operating the probe 2A and the tomographic image at the same time.

<Acquisition of ellipse center point and presentation of virtual fault plane>
In the present invention, the wall contour of the left ventricle LV extracted by the above-described method and the virtual tomographic plane corresponding to the position and angle of the probe 2A acquired by the ARToolkit are simultaneously configured by CG and superimposed on the real image in real time. Can be displayed.

  The operator of the probe 2A is instructed in advance so that the subject's chest is slowly traced by the probe 2A, and from the acquired tomogram, an elliptic center point distribution having the same phase is always measured by electrocardiogram synchronization.

  After about 10 ellipse center points are acquired, the position of the probe 2A is guided to the cross section including as many points as possible in the following manner. First, the average value of the distance between the tomographic plane at that time and the center point of each ellipse is obtained, and color presentation according to that value is performed.

FIG. 23 shows the positional relationship between an arbitrary ellipse center point P _i and the ultrasonic tomographic plane. A calculation formula for the distance L _i is shown in Formula 8. However, ΔABC is any three points on the tomographic plane, and n is a unit normal vector with respect to ΔABC. After that, an average value L _ave of all L _i was calculated, and the color of the tomographic plane was changed according to the setting of the tablet. As a result, the positional relationship between the distribution of each ellipse center point and the virtual tomographic plane can be visually grasped, and the position and angle of the probe 2A are adjusted accordingly.

<Acquisition of long-axis cross section>
Using this software, the estimation and presentation of the long-axis cross section were performed using 8 healthy males in their 20s as subjects. Results obtained from one of them are shown in FIG. Here, the ultrasonic tomographic image A is one tomographic image of the obtained six cross-sections in the short axis direction. The ultrasonic tomographic image B is a tomographic image obtained by approximating these to an ellipse and minimizing the distance between the center point and the tomographic plane L.

  Furthermore, the approximate ellipse and the three-dimensional positional relationship between the position and orientation of the probe 2A from which the tomographic image B is obtained are also shown. This tomographic image B satisfies the characteristics of the long-axis cross section mentioned above, and similar tomographic images were obtained for other subjects.

  In addition, for each of the eight subjects, the distance from the straight line to each ellipse center point and its standard deviation were calculated by approximating the obtained ellipse center point distribution to a straight line by the least square method. The results shown were obtained. Here, the error bar represents a range obtained by adding the distance average and its standard deviation.

  From FIG. 25, the distance average value is within 10 [mm] for any subject, and the value obtained by adding the standard deviation to the distance average is within 13 [mm]. Based on the above results, the long-axis cross section of the left ventricle LV as shown in FIG. 3 can be obtained only by fine adjustment even when applied to remote diagnosis by capturing a tomographic image so as to include a plurality of ellipse center points in the same cross section. Is easily obtained.

DESCRIPTION OF SYMBOLS 1 ... Ventricular wall information extraction apparatus 1A ... Vector measurement part 1B ... Initial search point setting part 1C ... Contour point extraction part 1D ... Ellipse approximation means 1E ... Three-dimensional measurement means 2 ... Ultrasonic diagnostic apparatus 2A ... Probe 2B ... Image generation part 2C ... Monitor LV ... Left ventricle MV ... Mitral valve

Claims (2)

An apparatus for extracting information on a ventricular wall based on an ultrasonic tomographic image of the heart imaged by a probe of an ultrasonic diagnostic apparatus,
An ultrasonic tomographic image is divided into a plurality of small regions, an average vector in each small region is defined as a region motion vector f of each small region, and the motion vector calculation unit calculates the following equation (9) , A motion vector measuring means for measuring a plurality of motion vectors;

(V _n is the calculated vector, N is the number of vectors in the small area)
Initial search point setting means for setting an intersection region of extension lines of a plurality of motion vectors measured by the motion vector measuring means as an initial search point;
A ventricular wall extracting device comprising: contour point extracting means for extracting a plurality of contour points of a ventricular wall by polar coordinates having the initial search point set by the initial search point setting means as an origin. Ellipse approximating means for drawing an approximate ellipse along the contour of the ventricular wall from a plurality of contour points of the ventricular wall extracted by the contour point extracting means;
The apparatus according to claim 1, further comprising: a three-dimensional measuring unit that obtains information on a shape and a three-dimensional position of a ventricular wall based on a plurality of approximate ellipses drawn according to a random operation of the probe. Ventricular wall information extraction device.