JPH0984793A - Ultrasonic image processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely detect the boundary of an intended tissue in ultrasonic image processing by setting learning windows on a plurality of mutually different texture areas, and setting a reference for calculating similarity from the distribution of texture characteristic quantities of the learning window within a characteristic space. SOLUTION: A probe control part 11 drives an ultrasonic probe 1 so as to perform the three-dimensional scanning of a subject. On the other hand, an ultrasonic observation device 12 successively forms a plurality of continued ultrasonic tomographic images from the echo signal of the subject, and successively outputs them to a tomographic image monitor 13 and a textural property diagnostic device 14. A learning window for gaining learning data of texture area is set on the ultrasonic tomographic image, and a texture characteristic quantity is calculated as learning data from the learning window. Further, a judgment window is moved on the ultrasonic tomographic image and successively set. In this case, learning windows are set on a plurality, of different texture areas. The reference for calculating similarity is set by the distribution of texture characteristic quantities within the characteristic space.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an ultrasonic image processing apparatus, and more particularly to ultrasonic image processing characterized by a portion for performing image analysis based on an image of an observed region obtained by three-dimensional scanning with an ultrasonic transducer. Regarding the device.

[0002]

2. Description of the Related Art Conventionally, as disclosed in Japanese Patent Laid-Open No. 6-30937, a three-dimensional scan combining two different scans such as a radial scan and a linear scan is performed.
An ultrasonic probe capable of obtaining three-dimensional echo data of a subject and a control device therefor are disclosed.

Further, Japanese Patent Application Laid-Open No. 7-47066 discloses that
Detects the surface of luminal organs such as the stomach, synthesizes an ultrasonic tomographic image and the surface of the shaded lumen on the same screen, and simultaneously observes surface irregularities and ultrasonic tone information useful for diagnosis A possible ultrasonic three-dimensional image processing is disclosed.

On the other hand, the tissue corresponding to the tissue of interest is extracted from the ultrasonic tomographic image by the texture analysis, and the ultrasonic image that enables the shape of the tissue and the state such as the tumor that cannot be clearly distinguished from the surrounding tissue can be identified. Processing devices are known.

Ultrasonic image processing devices based on texture analysis include, for example, Japanese Patent Publication No. 7-32777 and Japanese Patent Publication No.
The device disclosed in JP-A-34798 sets a region of interest for extracting a texture pattern of a target tissue,
The texture feature amounts of both areas are compared by sequentially setting the areas called search areas while moving them. Then, it is configured to compare the degree of similarity between the search area portion and the tissue of interest. Further, in this example, a process of three-dimensionally extracting a tissue is also disclosed. Furthermore, in this example, it is also disclosed that the display state such as the display color is changed and displayed according to the degree of similarity.

The apparatus disclosed in Japanese Patent Laid-Open No. 4-236952 is also configured to sequentially scan a region called a region of interest and extract the texture feature amount therein to perform texture analysis. In this device, in order to enable image analysis corresponding to ultrasonic resolution, S / N ratio, STC, etc., the image analysis parameters and the size of the region of interest are changed according to the ultrasonic wave propagation direction of the probe. Then, the texture analysis is performed. Further, the region of interest of this device moves in an annular shape on the image plane so as to move horizontally and vertically with respect to the ultrasonic wave propagation direction. By changing the size of the region of interest in this manner, when the tissue of interest such as a tumor spreads in a small region, the texture can be extracted so as not to mix with other texture features.

[0007]

However, an ordinary living tissue, such as a stomach tissue, is often composed of a plurality of tissues having different textures such as a layered structure. In addition, the echo inside the cancer tissue may not be uniform.

For example, consider that tissues A, B, and C exist on an ultrasonic tomographic image as shown in FIG.
Here, for example, the tissue A represents a tumor.

Japanese Patent Publication 7-32777, Japanese Patent Publication 7-
In the device disclosed in Japanese Patent No. 34798, since the region of interest can be set only by one type of tissue, the tissue B has a texture similar to that of the tumor (tissue A),
If the tissue C has a texture that is completely different from that of the tumor, the tissue B may be determined to be a tumor when the tissue B is determined based on the criteria determined for the tumor and the tissue C.

Further, since this apparatus is not configured to store the once calculated similarity, re-calculation is required when an image whose display state is changed according to the similarity is to be displayed again. In particular, the data for forming the three-dimensional image has a much larger data amount than the data for the two-dimensional image, and it takes much longer time to calculate the similarity. Therefore, there is a problem that it is necessary to wait for a very long time before displaying again.

Further, in this apparatus, since the range within the image which is sequentially set while the search region is moving is not designated, the process slows down the process such as extracting a texture of an unnecessary portion within the image. There are also problems.

In the device disclosed in Japanese Patent Laid-Open No. 4-236952, when the tissue of interest such as a tumor spreads in a small area, the area of interest must be made small. There is a problem that the number is small and it is not suitable for the process of extracting the texture feature amount.

The present invention has been made in view of the above circumstances, and provides an ultrasonic image processing apparatus capable of accurately and reliably detecting the surface of a tubular organ such as the stomach or the boundary of a target tissue such as a tumor. It is intended to be provided.

[0014]

An ultrasonic image processing apparatus according to the present invention is an ultrasonic probe for transmitting and receiving ultrasonic waves to and from an object to obtain echo data, and an ultrasonic tomographic image in the object based on the echo data. Ultrasonic tomographic image creating means for creating an image,
Storage means for storing the ultrasonic tomographic image, learning window setting means for setting a learning window for acquiring learning data of a texture region on the ultrasonic tomographic image, and texture feature amount as learning data from the learning window A learning data calculation means, a discrimination window setting means for sequentially setting the discrimination window on the ultrasonic tomographic image, and a discrimination data calculation means for calculating a texture feature amount as discrimination data from the discrimination window, Comparing the learning data and each of the discrimination data calculated by the discrimination data calculation means,
In an ultrasonic image processing apparatus including a comparison unit that calculates the degree of similarity between the texture region and the tissue corresponding to the discrimination window, the learning window setting unit sets the learning window on a plurality of different texture regions. However, the comparison means is provided with a similarity degree reference setting means for setting a reference for calculating the similarity degree based on the distribution of the texture feature amounts of the learning windows set on different texture regions in the feature space. .

In the ultrasonic image processing apparatus according to the present invention, the learning windows are set on a plurality of texture areas having different learning window setting means, and the similarity reference setting means are set on different texture areas. By setting the reference for calculating the similarity by the distribution of the texture feature amount of the learning window in the feature space, the boundary of the target tissue such as the surface of the tubular organ such as the stomach or the tumor is accurately and reliably detected. It is possible.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

1 to 9 relate to an embodiment of the present invention. FIG. 1 is a block diagram showing the configuration of an ultrasonic image processing apparatus, and FIG. 2 is a block diagram showing the configuration of the ultrasonic probe of FIG.
3 is a block diagram showing the configuration of the tissue characterization apparatus of FIG. 1, FIG. 4 is an explanatory view for explaining a window set by the mouse of FIG. 1, and FIG. 5 is an ultrasonic tomography obtained by the ultrasonic probe of FIG. 6 is an explanatory view for explaining the operation of the discrimination window moving circuit of FIG. 3, FIG. 7 is a first explanatory view for explaining the operation of the similarity criterion setting circuit of FIG. 3, and FIG. FIG. 9 is a second explanatory diagram for explaining the operation of the similarity reference setting circuit in FIG. 3, and FIG. 9 is a diagram showing an example of an ultrasonic three-dimensional image displayed on the analysis image monitor in FIG.

(Structure) In the ultrasonic image processing apparatus according to the present embodiment, as shown in FIG.
The ultrasonic transducer 3 that transmits and receives ultrasonic waves is attached to the tip of the flexible shaft 2, and is configured to cover the whole with a sheath 4.

As shown in FIG. 1, the signal lines of the flexible shaft 2 and the ultrasonic transducer 3 are connected to the ultrasonic probe controller 11. Then, the probe control unit 11
Outputs an echo signal from the ultrasonic transducer 3 to the ultrasonic observation device 12, the ultrasonic observation device 12 performs signal processing on the echo signal to generate an ultrasonic tomographic image, and the ultrasonic tomographic image is converted into an ultrasonic tomographic image. It is output to the image monitor 13 and the tissue property diagnostic device 14.

An input device such as a mouse 15 and a keyboard 16 is connected to the tissue property diagnostic device 14,
Based on the input data from the mouse 15 and the keyboard 16, the signal of the ultrasonic tomographic image from the ultrasonic observation device 12 is processed to detect the boundary of the target tissue.

The tissue characterization device 14 is an external storage device for storing an ultrasonic tomographic image obtained in the process of signal processing and a similarity image described later, for example, a hard disk device 17 and a magneto-optical disk device 18. Are connected, and an analysis image obtained by signal processing described later is displayed on the analysis image monitor 19.

The tissue texture diagnosing device 14 will be described in detail. As shown in FIG.
The ultrasonic tomographic image from the ultrasonic observation apparatus 12 is A / D converted by the A / D converter 21, and the A / D converter 21 outputs the ultrasonic tomographic image, which has become a digital signal, to the tomographic image storage circuit 22. The tomographic image storage circuit 22 stores the ultrasonic tomographic image.

The tomographic image storage circuit 22 includes a frame switch 2
3 are connected, and the frame switch 23 selectively outputs the frame of the ultrasonic tomographic image stored in the tomographic image storage circuit 22 to the frame memory 24. Further, this ultrasonic tomographic image is output to the zooming circuit 25 via the frame memory 24.

The magnification signal from the control circuit 26 is input to the zooming circuit 25, and the zooming circuit 25 performs the zooming processing of the ultrasonic tomographic image based on the magnification signal.

The output of the zooming circuit 25 is output to the input terminal C of the first switching device 27, and the output terminal O of the first switching device 27 analyzes the ultrasonic tomographic image to create an analysis image, which will be described later. It is connected to the analysis image creating circuit 28.

The analysis image forming circuit 28 includes an absolute position signal and control circuit 26 from an absolute position calculating circuit 34, which will be described later.
The learning / discrimination switching signal, the discrimination area designating signal, and the discrimination tomographic image designating signal are input, and the analysis image creating circuit 28 performs the analysis processing based on these signals to create the analysis image and also the analysis image storing circuit 29. To access the analysis image storage circuit 29 to store the analysis image. The analysis image created by the analysis image creation circuit 28 is the D / A converter 30.
And is D / A converted.

The analysis image output from the D / A converter 30 is output to the mixing amplification circuit 32 together with the synchronization signal from the synchronization signal generation circuit 31, and the analysis image is displayed on the analysis image monitor 19 by the mixing amplification circuit 32. To be done.

On the other hand, the mouse 15 and keyboard 1 described above
Reference numeral 6 is connected to the mouse control circuit 33 and the control circuit 26, respectively, and the mouse control circuit 33 receives the input from the mouse 15 and outputs the mouse movement signal to the absolute position calculation circuit 3
4 and receives the input from the keyboard 16, the control circuit 26 outputs the magnification signal, for example, to the zooming circuit 25.
And is output to the absolute position calculation circuit 34.
Then, the absolute position calculation circuit 34 calculates the absolute coordinates of a learning window described later on the ultrasonic tomographic image based on the mouse movement signal, and the absolute position calculation circuit 34 inputs the absolute coordinates to the input terminal I of the second switch 35 and the analysis image creation circuit 28. It is designed to output the coordinates.

The output terminal A of the second switching device 35 is connected to the input terminal A of the third switching device 36, and the output terminal O of the third switching device 36 is the characteristic amount calculating circuit 3 for calculating the texture characteristic amount.
Connected to 7. The output of the zooming circuit 25 and the mouse click signal from the mouse control circuit 33 are input to the feature amount calculation circuit 37, and the feature amount calculation circuit 37 outputs the texture feature amount to the input terminal I of the fourth switch 38. It is supposed to do.

By the way, the control circuit 2 is connected to the switching terminals X of the second switching device 35, the third switching device 36 and the fourth switching device 38.
A learning / discrimination switching signal corresponding to learning processing and discrimination processing, which will be described later, is inputted from 6, and each switching device switches the input or output to either of terminals A and B by the learning / discrimination switching signal. It is like this.

The output terminal A of the fourth switch 38 is a learning data storage circuit for storing texture feature amounts as learning data obtained in a learning window on an ultrasonic tomographic image set by the mouse 15 by a learning process described later. It is connected to 39. Further, an average value / variance value calculation circuit 40 is also connected to the learning data storage circuit 39, and the average value / variance value calculation circuit 40 has an average value of learning data stored in the learning data storage circuit 39 / The average value / variance value calculation start signal from the control circuit 26 for starting the calculation of the variance value is input.

The output terminal B of the second switch 35 is connected to the discrimination window moving circuit 41, and this discrimination window moving circuit 41 is connected.
Is a discrimination area designating signal from the control circuit 26 set by the keyboard 16, which indicates the discrimination area on the ultrasonic tomographic image set by the mouse 15 for the discrimination processing described later.
The discriminant area storage circuit 42 stores the discriminant tomographic image designation signal and the mouse click signal from the mouse control circuit 33. Then, the discrimination window moving circuit 41
Outputs a discrimination window position signal, which will be described later, to the input terminal B of the third switch 36 and a frame switch signal to the frame switch 23 and the analysis image creating circuit 28.

On the other hand, the learning data storage circuit 39 is connected to a comparison circuit 44 having a similarity reference setting circuit 43 therein, and the comparison circuit 44 outputs the output value from the output terminal B of the fourth switch 38. The comparison value with the value obtained by accessing the learning data storage circuit 39 is output to the similarity image creating circuit 45. The similarity image creating circuit 45 creates a similarity image, which will be described later, and outputs the output terminal O to the input terminal C of the fifth switching unit 46 connected to the analysis image creating circuit 28.

The external storage device control circuit 47 is an interface for accessing the above-mentioned hard disk device 17 and magneto-optical disk device 18, and the first control terminal X of the external storage device control circuit 47 switches data from the control circuit 26. A signal, and a display tomographic image designation signal from the control circuit 26 is input to the second control terminal Y. A signal from the zooming circuit 25 is input to the tomographic image input terminal E of the external storage device control circuit 47, and a signal from the similarity image creating circuit 45 is input to the similarity image input terminal F. Further, the tomographic image output terminal e of the external storage device control circuit 47 is connected to the first switch 27.
, And the similarity image output terminal f is connected to the input terminal D of the fifth switching unit 46. A data switching signal is input from the control circuit 26 to the switching terminals X of the first switching device 27 and the fifth switching device 46, and each of them switches the input to one of the terminals C and D. It is like this.

(Operation) Next, the operation of the learning processing of the ultrasonic image processing apparatus configured as described above will be described.

As shown in FIG. 2, the ultrasonic probe control unit 11 combines the radial scan (FIG. 2A) and the linear scan (FIG. 2B) to perform a three-dimensional scan of the subject. Driving the ultrasonic probe 1 (Fig. 2
(C)).

As shown in FIG. 1, in the ultrasonic observation apparatus 12, the ultrasonic probe 1 is passed through the ultrasonic probe controller 11.
A plurality of continuous ultrasonic tomographic images are sequentially created from the echo signals of the subject obtained by the three-dimensional scanning of 3 and are sequentially output to the tomographic image monitor 13 and the tissue property diagnostic apparatus 14. Then, the tomographic image monitor 13 sequentially displays the ultrasonic tomographic images.

In the tissue property diagnosing device 14, as shown in FIG. 3, the tomographic image storage circuit 22 stores continuous, for example, 100 ultrasonic tomographic images A / D converted by the A / D converter 21. To do. The frame switch 23 selects one of the plurality of ultrasonic tomographic images designated by the frame switching signal. The frame memory 24 stores the selected ultrasonic tomographic image. The zooming circuit 25 magnifies the ultrasonic tomographic image selected by the frame memory at the magnification designated by the magnification signal, and outputs it to the input terminal C of the first switching device 27.
This magnification is set by the keyboard 16.

The first switch 27 switches the input to the input terminal C when the content of the data switching signal indicates "live data". Therefore, the enlarged ultrasonic tomographic image is input to the analysis image creating circuit 28. The contents of this data switching signal are set by the keyboard 16.

On the other hand, when the operator moves the mouse 15, the mouse control circuit 33 outputs the moving distance information as a mouse moving signal. The absolute position calculation circuit 34 calculates the absolute coordinates of the learning window on the ultrasonic tomographic image from the mouse movement signal,
The coordinates are converted by the magnification information based on the magnification signal.

As shown in FIG. 4, it is assumed that the learning window has a fan shape corresponding to the distance from the ultrasonic transducer 3 on the ultrasonic tomographic image. At this time, the position of the learning window indicated by the absolute position signal is represented by the pixel closest to the ultrasonic transducer 3 on the fan-shaped center line as shown by the pixel P.

Then, information on the position and orientation of the learning window is output as an absolute position signal so that the learning window whose position is indicated by the absolute coordinates directly faces the sound ray from the ultrasonic transducer 3.
On the other hand, when the content of the learning / discrimination switching signal indicates “learning”, the analysis image creating circuit 28 superimposes the absolute position signal and the ultrasonic tomographic image magnified by the zooming circuit 25 and outputs it. . By the way, this learning / discrimination switching signal is set by the keyboard 16.

The D / A converter 30 converts the superimposed image into D / A.
Convert. The mixing / amplifying circuit 32 mixes the D / A converted image with the sync signal generated by the sync signal generating circuit 31 so that the image can be output to the analysis image monitor 19. The analysis image monitor 19 displays the mixed superimposed image.

In this way, the operator can move the learning window on the enlarged ultrasonic tomographic image.

A method for setting a plurality of learning windows in different regions on the ultrasonic tomographic image enlarged as shown in FIG. 5 will be described below. In FIG. 5, it is considered that the tissue A, the tissue B, and the tissue C are present on the ultrasonic tomographic image.

The second switch 35 and the third switch 36 learn /
When the content of the discrimination switching signal indicates "learning", the input or output is set to A. For this reason,
The absolute position signal is directly output to the characteristic amount calculation circuit 37.

The feature amount calculation circuit 37 refers to the enlarged image enlarged by the zooming circuit 25 every time the mouse click signal is input, and the feature amount calculation circuit 37 sets the learning position within the learning window set by the absolute position signal of the enlarged image. The texture feature amount is calculated.
The content of the learning / discrimination switching signal of the fourth switching device 38 is "learning".
If it is to instruct, the output is set to the output terminal A. Each time the mouse click signal is input, the texture feature amount is written in the learning data storage circuit 39 by changing the address.

By the way, this mouse click signal is transmitted through the mouse control circuit 33 when the mouse 15 is clicked. Therefore, every time the mouse 15 is moved and clicked, the texture feature amount of the learning window corresponding to the coordinates is calculated and written in the learning data storage circuit 39.

The average value / variance value calculation circuit 40
When the variance value calculation signal is input, the learning data storage circuit 39
The texture feature amount stored in is read out, and the average value and the variance value thereof are calculated. When a plurality of texture feature quantities are used, the covariance value of each is also calculated. The calculated mean value, variance value, and covariance value are stored in the learning data storage circuit 39 again with different addresses. The average value / variance value calculation signal is received by the keyboard 16 and transmitted.

As described above, by repeating the click of the mouse 15 and the operation of the keyboard 16, the learning window of the organization A as shown in FIG. 5 is set. Then, the average value, the variance value, and the covariance value of the texture feature amount in the learning window are calculated and written in the learning data storage circuit 39. Further, by repeating the same operation for the tissue B, the tissue C, and the lumen, the average value, the variance value, and the covariance value of each texture feature amount are written in the learning data storage circuit.

Next, the discrimination operation will be described.

When a series of learning operations are completed, the operator switches the learning / discrimination switching signal to the content for instructing "discrimination" via the keyboard 16. Then, the second switch 35 switches the output to the output terminal B. Therefore, the information on the absolute coordinates and orientation of the window on the ultrasonic tomographic image calculated by the absolute position calculation circuit 34 as the absolute position signal from the mouse movement signal is output to the discrimination window movement circuit 41.

When the content of the discrimination area designation signal indicates "during discrimination area designation", the discrimination window moving circuit 41 inputs this absolute coordinate and orientation information simultaneously with the input of the mouse click signal. Write to 42. This discrimination area designation signal is set by the keyboard 16.

By the way, the analysis image forming circuit 28 indicates the position by the enlarged ultrasonic tomographic image and the absolute position signal when the content of the discrimination area designation signal is "during discrimination area designation" as in the "learning" operation. Since the triangular cursors to be output are superimposed and output, it is possible to specify the discrimination region on the analysis image monitor 19 on the ultrasonic tomographic image enlarged as shown in FIG. The cursor is shown in FIG. 6 as a triangular indicator. Here, on the analysis image monitor 19, the operator moves the cursor to the points indicated by points P, Q, and R in FIG. 6 and clicks the mouse 15 to perform a discrimination process of extracting a tumor. The discrimination area is stored in the discrimination area storage circuit 42.

The discrimination window moving circuit 41 is the discrimination area storage circuit 4
With reference to the discrimination area stored in 2, the discrimination window position signal is sequentially updated and output while adding an appropriate number. Then, when the coordinates of the discrimination window designated by the discrimination window position signal are about to protrude from the discrimination area, a reset is applied to correct. At this time, the discrimination window moving circuit 41 is set so that the discrimination window is always oriented so as to face the sound ray from the ultrasonic transducer 3.

In this way, in the discrimination area shown by the arrow in FIG. 6, scanning is realized so that the discrimination window always faces the sound ray from the ultrasonic transducer 3. Further, the discrimination window has a fan shape as shown in FIG. 4, like the learning window.

When the learning / discrimination switching signal indicates "discrimination", the third switching device 36 switches the input to the input terminal B, and therefore the discrimination window position signal is output to the characteristic amount calculating circuit 37. . The feature amount calculation circuit 37 refers to the enlarged image enlarged by the zooming circuit 25 as in the learning window, and the texture feature amount in the discrimination window whose position is indicated by the discrimination window position signal in the enlarged image. To calculate.
The fourth switch 38 switches the output to the output terminal B when the learning / discrimination switching signal has the content of instructing "discrimination". Therefore, the position of the discrimination window and the texture feature amount are output to the comparison circuit 44. To be done.

The comparison circuit 44 is a learning data storage circuit 39.
To read the average value, the variance value, and the covariance value of the texture feature amount for each texture region of the tissue A, the tissue B, the tissue C, and the lumen that have been written, and to determine the texture feature amount in the discrimination window. The degree of similarity between the two is calculated by comparison.

Details of the operation of the comparison circuit 44 will be described below. The similarity reference setting circuit 43 provided in the comparison circuit 44 calculates the average value of the texture feature amount for each texture region of the tissue A, the tissue B, the tissue C, and the lumen written in the learning data storage circuit 39. The distribution of the texture feature amount of each tissue predicted from the variance value and the covariance value is plotted in the feature space having these texture feature amounts in the orthogonal coordinates. This plot is shown in FIG.

In FIG. 7, as the texture feature quantity f ₁ ,
Three types of f ₂ and f ₃ are used, and for example, parameters that can quantitatively measure the texture in the window, such as the average brightness value in the window, the brightness dispersion value, and the depth of the unevenness of the brightness value are used. In FIG. 7, the points indicated by +, ○, Δ, and ◇ are points when the learning windows shown in FIG. 5 are plotted in the feature space, and these are collected for each tissue and distributed. Should do. The points labeled μ ^A , μ ^B , μ ^C , and μ (inside) represent the average value for each texture region of tissue A, tissue B, tissue C, and lumen, and the average The value / variance value calculation circuit 40 calculates the value by the following formula.

[0061]

[Equation 1] Here, μ ₁ ^A , μ ₂ ^A , μ ₃ ^A : average values of texture feature amounts f ₁ , f ₂ , f ₃ of the learning windows belonging to the tissue A, and μ ^A , μ ^B , μ ^C , μ ( The ellipsoid shown around (inside) indicates the dispersion value in an arbitrary direction.For example, the equation of the ellipsoid around μ ^A is described as follows, of which the covariance matrix Σ ^A is It is calculated by the average value / variance value calculation circuit 40.

^T (x−μ ^A ) (Σ ^A ) ⁻¹ (x−μ ^A ) = p where:

[Equation 2] ^t x = (transpose of x) = (f ₁ , f ₂ , f ₃ ) (Σ ^A ) ⁻¹ = inverse matrix of covariance matrix Σ ^A

(Equation 3) σ ^A _ii (i = 1,2,3): variance value of texture feature quantity f _i for organization A σ ^A _ij (i, j = 1,2,3): organization of texture feature quantities f _i and f _j A
Then, the similarity criterion setting circuit 43 pays attention to the tissues A and B and shades them in a direction in which both distributions are most separated. This direction is shown as d in FIG. Moreover, the oblique image to d of both distributions is shown in FIG. μ ^A 'and μ ^B ' are oblique images of μ ^A and μ ^B , respectively. The direction of d is approximated by the following formula.

D = {(Σ ^A + Σ ^B ) / 2} ⁻¹ (μ ^A −μ ^B ) Further, the similarity reference setting circuit 43 sets a threshold value α between both oblique images. α is a point that internally divides between μ ^A 'and μ ^B ' by the way of spreading the tail of the oblique image to d of the ellipsoid. The internal division ratio is represented by ^t dΣ ^A d: ^t dΣ ^B d.

Thereafter, the similarity reference setting circuit 43 plots the texture feature quantity of the discrimination window in the feature space. This plot is shown by x in FIG.

Thereafter, the comparison circuit 44 shades this X in the d direction. This is indicated by θ in FIG. The comparison circuit 44
The texture region to which the discrimination window belongs is discriminated depending on which side of μ ^A 'and μ ^B ' is this θ with α as the threshold value. In this case, x belongs to organization B.

Further, the comparison circuit 44 and the similarity reference setting circuit 43 repeat this series of processing for each combination of μ ^A , μ ^B , μ ^C and μ (in), and in the oblique image for each combination, It is determined which of the texture regions of the tissue A, the tissue B, the tissue C, and the lumen the determination window finally belongs to, based on which side has the greatest number of x.

Then, the comparison circuit 44 determines the position of the discrimination window and
The attributes of the texture regions of the lumen to which the discrimination window belongs, the tissue A, the tissue B, and the tissue C are output to the similarity image creation circuit 45.

The similarity image creating circuit 45 creates a similarity image having a different color according to the texture area to which the discrimination window belongs. When the data switching signal indicates "live data", the first switching device 27 and the fifth switching device 4
Since 6 switches the input to the input terminal C, the ultrasonic tomographic image enlarged by the zooming circuit 25 and the similarity image are output to the analysis image creating circuit 28. Analysis image creation circuit 28
Superimposes the enlarged ultrasonic tomographic image and the similarity image.

The discrimination window moving circuit 41 outputs a frame switching signal when the discrimination in the discrimination region of one ultrasonic tomographic image is completed. When the frame switching signal is input, the analysis image creation circuit 28 writes one superposed image in the analysis image storage circuit 29. Then, another magnified image designated by the discrimination tomographic image designation signal is discriminated in the same discrimination region. This discriminating tomographic image designation signal is generated by the operator using the keyboard 16 and is stored in the tomographic image storage circuit 22 in succession.
The number of images to be processed is set for each image.

By repeating a series of processing in this way, with respect to the enlarged ultrasonic image of a plurality of continuous ultrasonic tomographic images set by the operator using the keyboard 16, each similarity image is created and Superimpose the magnified image,
It is stored in the analysis image storage circuit 29.

The analysis image creating circuit 28 writes all the continuous ultrasonic tomographic images specified by the discriminating tomographic image specifying signal in the analysis image storage circuit 29, and then collectively reads them to perform light ray processing and hidden surface processing. After processing such as erasing,
Create a three-dimensional image. At this time, since the portion corresponding to the inner cavity is seen through, the data corresponding to the discrimination window discriminated as the inner cavity is omitted. The display colors of the data corresponding to the discrimination windows for discriminating the tissues A, B, and C are changed.
Following this, the analysis image creating circuit 28 and the D / A converter 3
The operation of 0, the synchronization signal generation circuit 31, the mixing amplification circuit 32, and the analysis image monitor 19 is as described above. In this manner, for example, as shown in FIG. 9, the display color is changed according to the degree of similarity with the tissue A, the tissue B, and the tissue C, and an ultrasonic three-dimensional image through the lumen is displayed on the analysis image monitor 19. Is output to.

Next, the operation of the external storage device 47 will be described.

When the data switching signal indicates "live data", the external storage device control circuit 47 inputs from the tomographic image input terminal E and the similarity image input terminal F, and the zooming circuit 25 sequentially enlarges the data. The plurality of continuous ultrasonic tomographic images and the similarity image created by the similarity image creating circuit 45 are written in the hard disk device 17 and the magneto-optical disk device 18. That is, in the hard disk device 17 and the magneto-optical disk device 18, 100 continuous ultrasonic tomographic images and the similarity calculated for each discrimination window sequentially set on the ultrasonic tomographic images are calculated for each coordinate. Will be stored in.

On the other hand, when the data switching signal indicates "three-dimensional load data", the external storage device control circuit 47 causes the hard disk device 17 and the magneto-optical disk device 18 to store a plurality of continuous ultrasonic tomographic images. The image and the similarity image corresponding thereto are read out and output from the tomographic image output terminal e and the similarity image output terminal f. When the data switching signal indicates "three-dimensional load data", the first switching device 27 and the fifth switching device 46 switch the input to the input terminal D. Therefore, the read ultrasonic tomographic image and the similarity image are output to the analysis image creating circuit 28. The analysis image creating circuit 28 creates an ultrasonic three-dimensional image as shown in FIG. 9 when a plurality of ultrasonic tomographic images are input. Following this, the D / A converter 30, the synchronization signal generation circuit 31, the mixing amplification circuit 32, and the analysis image monitor 19
Is the same as described above, and similarly, tissue A, tissue B,
Ultrasound 3 whose display color is changed according to the degree of similarity to the tissue C
The three-dimensional image is output on the analysis image monitor 19.

Further, when the data switching signal has the content of instructing "two-dimensional load data", the external storage device control circuit 47 designates the display tomographic image among the ultrasonic tomographic images to the hard disk device 17 and the magneto-optical disk device 18. One ultrasonic tomographic image designated by the signal and the similarity image corresponding thereto are read out and output from the tomographic image output terminal e and the similarity image output terminal f. The contents of this display tomographic image designation signal are set by the keyboard 16. Data switching signal is "2"
When the content indicates "dimensional load data", the first switching device 27 and the fifth switching device 46 switch the input to the input terminal D. Therefore, the read ultrasonic tomographic image and the similarity image are analyzed image creation. The analysis image creating circuit 28 superimposes the ultrasonic tomographic image and the similarity image corresponding to the ultrasonic tomographic image when the number of input ultrasonic tomographic images is one. The operations of the converter 30, the synchronization signal generation circuit 31, the mixing amplification circuit 32, and the analysis image monitor 19 are as described above. Similarly, the tissue A, the tissue B, and the tissue C
An ultrasonic tomographic image whose display color is changed according to the degree of similarity with is output on the analysis image monitor 19.

(Effect) As described above, the mouse 15 sets a plurality of learning windows, the texture feature amounts of the learning windows set on different texture regions are plotted on the feature space, and the texture feature amounts of the space are plotted. Since the comparison circuit 44 includes the similarity reference setting circuit 43 that sets the reference of the similarity according to the distribution, the textures in the image such as the lumen, the tissue A, the tissue B, and the tissue C are different for each region. When divided into, you can learn by distinguishing these. FIG.
As shown in, for example, even if the tissue A represents a tumor and the tissue B has a texture similar to that of the tumor (tissue A) and the tissue C has a texture completely different from that of the tumor, the tissue B is regarded as the tumor. The possibility of making a judgment is significantly reduced as compared with the conventional technique, and the accuracy of detecting the surface of a tubular organ such as the stomach or the boundary of a target tissue such as a tumor is significantly improved.

Further, a plurality of learning windows are set in one kind of organization, and the learning data and the discrimination data are compared with each other based on the standard calculated from the statistical values such as the average value and the variance value of the texture feature amount. Then, since the similarity between the texture region and the tissue corresponding to the discrimination window is compared, the texture feature amount generated during the learning operation in each of the texture regions of the tissue A, the tissue B, the tissue C, and the lumen. The similarity can be calculated by suppressing the variation of. Therefore, the accuracy of detecting the boundary of the target tissue is also improved.

Further, the zooming circuit 25 for enlarging the ultrasonic tomographic image is provided, and the learning window and the discrimination window are set from the enlarged ultrasonic tomographic image. Above all, a plurality of texture regions are not mixed in the learning window, the learning window and the discrimination window can be set without reducing the number of pixels, and the accuracy of detecting the boundary of the target tissue is improved.

Further, a learning window is set for a specific image of the continuous ultrasonic tomographic images, and a discrimination window is set for each of the continuous ultrasonic tomographic images including the specific ultrasonic tomographic image. Since the objects are moved and sequentially set to be discriminated, the lumen can be detected easily even with a plurality of ultrasonic tomographic images.
Therefore, as shown in FIG. 9, the surface of a luminal organ such as the stomach is detected, and an ultrasonic tomographic image and a shaded surface of the luminal surface are combined on the same screen to form surface irregularities and ultrasonic waves useful for diagnosis. It is possible to easily create an ultrasonic three-dimensional image in which the gradation information of 3) can be observed simultaneously. 3 consecutive ultrasonic tomographic images
The images were taken by changing the positions little by little by the dimensional scan, and the tissues A, B, C shown in FIG.
The texture feature amount of a tissue such as a lumen is similar in each ultrasonic tomographic image. Therefore, even if the other ultrasonic tomographic images of the plurality of continuous ultrasonic tomographic images are discriminated by using the learning data thus obtained in the specific image, the discrimination can be performed with good results.

Further, a hard disk 35, a magneto-optical disk 36 for storing the degree of similarity with the learned data for each coordinate of the discrimination window, and an external storage device control circuit 34 for controlling these are provided, and depending on the stored degree of similarity. Since the analysis monitor 19 for changing the display state of the ultrasonic tomographic image is provided, it is not necessary to recalculate the similarity when redisplaying the similarity, and the similarity with the target tissue can be displayed at high speed. it can.

Further, since the discrimination window moving circuit 41 for designating the region in the ultrasonic tomographic image in which the discrimination window is sequentially moved and set is provided, the discrimination window does not extract the texture of the unnecessary portion in the image. The process of detecting the boundary of the tissue of interest can be efficiently performed.

Further, in the present embodiment, the control circuit 26 changes the contents of the display tomographic image designating signal and the data switching signal and outputs the changed signals to the external storage device control circuit 47, and the ultrasonic image input by the analysis image forming circuit 28. Depending on whether the tomographic image is plural or one, a plurality of continuous ultrasonic tomographic images are superimposed on the similarity image to output an ultrasonic three-dimensional image, or one ultrasonic tomographic image and the corresponding degree of similarity. Since the processing for superimposing and outputting the images is divided, only by using the ultrasonic image processing apparatus according to the present embodiment, the degree of similarity with the texture region is observed in both two-dimensional and three-dimensional ultrasonic images. be able to.

Further, as shown in FIGS. 5 and 6, since the learning window and the discrimination window are scanned and set so as to directly face the sound ray from the ultrasonic transducer, the texture feature is set under the same condition at any position. The amount can be calculated. Further, as shown in FIG. 4, since the learning window and the discrimination window are fan-shaped, it is possible to calculate the texture feature amount according to the characteristics of the ultrasonic beam in which the beam becomes wider as the distance increases.

(Modification) Although the mouse is used to set the learning window and the discrimination window, the present invention is not limited to this, and another indicator such as a trackball may be used.

Although three types of texture feature amounts are used, any number of other types may be used, and one type or two types may be used. Further, any parameter may be used as long as it is a parameter representative of the texture of the image.

Furthermore, although p = 1 in the equation of the ellipsoid, there is no problem if p> 0. Further, the internal division ratio of the internal division point α may be, for example, the midpoint of μ ^A 'and μ ^B ' without using dispersion.

Further, the color arrangement assigned to the feature space may be any other color than the one described in this embodiment. Further, another method may be used as long as it is a method of changing not only the color but also the display state such as brightness and saturation.

The external storage device used is, for example,
A recording device other than a hard disk or a magneto-optical disk such as a writable CD-ROM can be used.

Further, the points representing the positions of the learning window and the discrimination window may not be represented by the pixel closest to the ultrasonic transducer on the fan-shaped center line as described above. For example, it may be one of the four corners of the fan-shaped learning window.

[Additional Notes] (Additional Item 1) An ultrasonic probe for transmitting and receiving ultrasonic waves to and from an object to obtain echo data, and an ultrasonic tomographic image for creating an ultrasonic tomographic image in the object from the echo data. Creating means, storage means for storing the ultrasonic tomographic image, learning window setting means for setting a learning window for acquiring learning data of a texture region on the ultrasonic tomographic image, and learning data from the learning window As learning data calculation means for calculating texture feature amount, discrimination window setting means for sequentially setting the discrimination window on the ultrasonic tomographic image, and discrimination data for calculating texture feature amount as discrimination data from the discrimination window A calculation unit, the learning data, and each of the discrimination data calculated by the discrimination data calculation unit are compared with each other to correspond to the texture region and the discrimination window. In an ultrasonic image processing apparatus including a comparing unit that calculates a similarity with a texture, the learning window setting unit sets the learning windows on a plurality of different texture regions, and the comparing units have different texture regions. An ultrasonic image processing apparatus, comprising: similarity reference setting means for setting a reference for calculating a similarity based on a distribution of texture feature amounts of the learning window set above in a feature space.

(Additional Item 2) An ultrasonic probe for transmitting / receiving ultrasonic waves to / from an object to obtain echo data, and an ultrasonic tomographic image creating means for creating an ultrasonic tomographic image of the object from the echo data. Storage means for storing the ultrasonic tomographic image, learning window setting means for setting a learning window for acquiring learning data of a texture region on the ultrasonic tomographic image, and texture characteristics as learning data from the learning window Learning data calculation means for calculating the amount, discrimination window setting means for moving the discrimination window on the ultrasonic tomographic image and sequentially setting the discrimination window,
The texture data and the discrimination window are compared by comparing the learning data and each of the discrimination data calculated by the discrimination data calculation unit with the discrimination data calculation unit that calculates the texture feature amount as the discrimination data from the discrimination window. In the ultrasonic image processing apparatus including a comparison means for calculating the degree of similarity with the tissue corresponding to the ultrasonic probe, the ultrasonic probe is provided so that the ultrasonic probe can obtain the three-dimensional echo data of the subject. The ultrasonic tomographic image creating unit creates a plurality of continuous ultrasonic tomographic images, the storage unit stores the plurality of ultrasonic tomographic images, and the learning window setting unit includes a three-dimensional control unit for controlling the ultrasonic tomographic images. Setting the learning window for a specific ultrasonic tomographic image among the plurality of ultrasonic tomographic images,
The discrimination window setting means moves the discrimination window for each of a plurality of continuous ultrasonic tomographic images including the specific ultrasonic tomographic image for which the learning window is set, and sequentially sets the discrimination window. Ultrasonic image processing device.

(Additional Item 3) The learning window is set on a plurality of the texture areas having different learning window setting means,
The comparison means includes a similarity reference setting means for setting a reference for calculating a similarity based on a distribution in the feature space of texture feature amounts of the learning windows set on different texture regions. The ultrasonic image processing apparatus according to item 2.

(Additional Item 4) The similarity criterion setting means includes a statistical value calculating means for calculating a statistical value of the learning data calculated for each learning window by the learning data calculating means, and the statistical value The ultrasonic image processing apparatus according to appendix 1 or 3, wherein a criterion for calculating the degree of similarity is set.

(Additional Item 5) The similarity criterion setting means calculates the distribution in the feature space of the texture feature amount of the learning window set on two texture regions of the plurality of different texture regions, Note 4 characterized in that the two distributions are shaded in the direction most separated and are repeated for other combinations of texture regions.
The ultrasonic image processing device according to item 1.

(Additional Item 6) An enlarging means for enlarging the ultrasonic tomographic image is provided, the learning window setting means sets the learning window from the enlarged ultrasonic tomographic image, and the discrimination window setting means The ultrasonic image processing apparatus according to any one of appendices 1, 2, 3, 4 or 5, wherein the discrimination window is set from the enlarged ultrasonic tomographic image.

(Additional Item 7) According to the similarity stored in the similarity storage means for storing the similarity for each coordinate of the discrimination window that is sequentially moved and set, and the similarity stored in the similarity storage means, 7. An ultrasonic image processing apparatus according to any one of appendices 1, 2, 3, 4, 5 or 6, further comprising: a display unit that changes and displays a display state of an ultrasonic tomographic image. .

(Additional Item 8) Additional Items 1, 2, 3, characterized in that a determination region designating unit for designating a region in the ultrasonic tomographic image in which the determination window is sequentially moved and set is provided.
The ultrasonic image processing device according to any one of 4, 5, 6 or 7.

(Supplementary Note 9) An ultrasonic three-dimensional image is created from a plurality of ultrasonic tomographic images and similarity storing means for storing the similarity for each coordinate of the discrimination window which is sequentially moved and set. Supplementary note 2 characterized by comprising: three-dimensional processing means and display means for changing and displaying the display state of the ultrasonic three-dimensional image according to the similarity stored by the similarity storage means. Alternatively, the ultrasonic image processing apparatus according to Item 3.

(Additional Item 10) Three-dimensional discrimination area designating means for designating an ultrasonic tomographic image in a plurality of continuous ultrasonic tomographic images or a region in the ultrasonic tomographic image in which the discrimination windows are sequentially moved and set. Additional items 2 and 3 characterized in that
Or the ultrasonic image processing device according to any one of 9.

An object of Supplements 7 and 9 of the present invention is to provide an ultrasonic image processing apparatus capable of displaying the degree of similarity to the tissue of interest at high speed.

It is an object of the additional items 8 and 10 of the present invention to provide an ultrasonic image processing apparatus for efficiently performing the process of detecting the boundary of the target tissue.

The operation of the ultrasonic image processing apparatus according to supplementary note 2 of the present invention is that the ultrasonic probe transmits and receives ultrasonic waves to and from the subject to obtain echo data. The three-dimensional control means controls the ultrasonic probe so that the ultrasonic probe can obtain three-dimensional echo data of the subject.

The ultrasonic tomographic image creating means creates a plurality of continuous ultrasonic tomographic images in the subject from the echo data.

The storage means stores a plurality of ultrasonic tomographic images. The learning window setting means sets a learning window for acquiring learning data of a texture region on an ultrasonic tomographic image for a specific ultrasonic tomographic image among a plurality of ultrasonic tomographic images.

The learning data calculating means calculates the texture feature amount as learning data from the learning window. The discrimination window setting means moves the discrimination window for each of a plurality of continuous ultrasonic tomographic images including the specific ultrasonic tomographic image for which the learning window is set, and sequentially sets the discrimination windows.

The discrimination data calculating means calculates the texture feature amount as discrimination data from the discrimination window.

The comparing means compares the learning data with each of the discrimination data calculated by the discrimination data calculating means to calculate the degree of similarity between the texture region and the tissue corresponding to the discrimination window.

The operation of the ultrasonic image processing apparatus according to supplementary note 3 of the present invention is that the learning window setting means sets a learning window on a plurality of different texture regions. The similarity criterion setting means is
A criterion for calculating the degree of similarity is set based on the distribution of the texture feature amounts of the learning windows set on different texture regions in the feature space.

The comparing means compares the learning data with each of the discrimination data calculated by the discrimination data calculating means, and compares the texture area and the tissue corresponding to the discrimination window with the reference set by the similarity reference setting means. Calculate the similarity. Other operations are the same as those of the ultrasonic image processing apparatus described in appendix 2.

In the operation of the ultrasonic image processing apparatus according to supplementary note 4 of the present invention, the statistical value calculating means calculates the statistical value of the learning data calculated for each learning window by the learning data calculating means. The similarity criterion setting means sets a criterion for calculating the similarity from this statistical value. Other functions are the same as those of the ultrasonic image processing apparatus described in appendices 1 and 3.

The operation of the ultrasonic image processing apparatus according to supplementary note 6 of the present invention is such that the magnifying means magnifies the ultrasonic tomographic image. The learning window setting means sets the learning window from the enlarged ultrasonic tomographic image. The discrimination window setting means sets the discrimination window from the enlarged ultrasonic tomographic image. Other functions are described in additional items 1 and 2,
This is the same as the ultrasonic image processing device described in 3 and 4.

In the operation of the ultrasonic image processing apparatus according to supplementary note 7 of the present invention, the similarity storage means stores the similarity for each coordinate of the discrimination window which is sequentially moved and set.

The display means changes and displays the display state of the ultrasonic tomographic image according to the similarity stored by the similarity storage means. Other functions are described in additional items 1, 2, 3, 4,
This is the same as the ultrasonic image processing device described in No. 6.

The operation of the ultrasonic image processing apparatus according to supplementary note 8 of the present invention is that the discrimination area designating means designates an area in the ultrasonic tomographic image in which the discrimination windows are sequentially moved and set. Other operations are the same as those of the ultrasonic image processing apparatus described in the additional items 1, 2, 3, 4, 6, and 7.

The operation of the ultrasonic image processing apparatus according to claim 9 of the present invention is that the three-dimensional processing means creates an ultrasonic three-dimensional image from a plurality of ultrasonic tomographic images. The display unit displays the ultrasonic wave 3 according to the similarity stored by the similarity storage unit.
The display state of the three-dimensional image is changed and displayed. Other operations are the same as those of the ultrasonic image processing apparatus described in the additional items 2 and 3.

The operation of the ultrasonic image processing apparatus according to supplementary note 10 of the present invention is that the three-dimensional discriminant area designating means is configured so that the discriminating windows are sequentially moved and set in a plurality of consecutive ultrasonic tomographic images. Designate an area in the image or ultrasonic tomographic image.
Other operations are the same as those of the ultrasonic image processing apparatus described in the additional items 2, 3, and 9.

[0117]

As described above, according to the ultrasonic image processing apparatus of the present invention, the learning windows are set on a plurality of texture areas having different learning window setting means, and the similarity reference setting means are provided on different texture areas. The criterion for calculating the similarity is set by the distribution of the texture feature amount of the learning window set in the feature space in the feature space, so that the boundary of the tissue of interest such as the surface of the tubular organ such as the stomach or the tumor can be accurately and reliably The effect is that it can be detected.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a configuration of an ultrasonic image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a configuration diagram showing a configuration of the ultrasonic probe of FIG.

FIG. 3 is a block diagram showing the configuration of the tissue property diagnosis apparatus of FIG.

FIG. 4 is an explanatory diagram illustrating a window set by the mouse of FIG.

5 is an explanatory diagram illustrating an ultrasonic tomographic image obtained by the ultrasonic probe of FIG.

6 is an explanatory view for explaining the operation of the discrimination window moving circuit of FIG.

FIG. 7 is a first explanatory diagram illustrating an operation of the similarity reference setting circuit in FIG.

FIG. 8 is a second explanatory diagram illustrating the operation of the similarity reference setting circuit in FIG.

9 is a diagram showing an example of an ultrasonic three-dimensional image displayed on the analysis monitor of FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Ultrasonic probe 2 ... Flexible shaft 3 ... Ultrasonic transducer 4 ... Sheath 11 ... Probe control part 12 ... Ultrasonic observation device 13 ... Ultrasonic tomographic image monitor 14 ... Tissue property diagnostic device 15 ... Mouse 16 ... Keyboard 17 ... Hard disk device 18 ... Magneto-optical disk device 19 ... Analysis image monitor 21 ... A / D converter 22 ... Tomographic image storage circuit 23 ... Frame switching device 24 ... Frame memory 25 ... Zooming circuit 26 ... Control circuit 27 ... First switching device 28 ... analysis image creation circuit 29 ... analysis image storage circuit 30 ... D / A converter 31 ... synchronization signal generation circuit 32 ... mixing amplification circuit 33 ... mouse control circuit 34 ... absolute position calculation circuit 35 ... second switching device 36 ... third Switching device 37 ... Feature amount calculation circuit 38 ... Fourth switching device 39 ... Learning data storage circuit 40 ... Average value / dispersion value calculation circuit 1 ... discrimination window moving circuit 42 ... judgment area storage circuit 43 ... similarity reference setting circuit 44 ... comparator circuit 45 ... similarity image creation circuit 46 ... fifth switching unit 47 ... external storage device control circuit

Claims (1)

[Claims] 1. An ultrasonic probe for transmitting and receiving ultrasonic waves to and from an object to obtain echo data, an ultrasonic tomographic image creating unit for creating an ultrasonic tomographic image of the object from the echo data, Storage means for storing an acoustic tomographic image; learning window setting means for setting a learning window for obtaining learning data of a texture region on the ultrasonic tomographic image; and texture feature amount calculation as learning data from the learning window. Learning data calculation means, a discrimination window setting means for moving the discrimination window on the ultrasonic tomographic image and sequentially setting the discrimination window, a discrimination data calculation means for calculating a texture feature amount as discrimination data from the discrimination window, and the learning The data and each of the discrimination data calculated by the discrimination data calculating means are compared to determine the similarity between the texture region and the tissue corresponding to the discrimination window. In an ultrasonic image processing apparatus having a comparing means for outputting, the learning window setting means sets the learning window on a plurality of different texture areas, and the comparing means sets the texture areas on different texture areas. An ultrasonic image processing apparatus comprising: a similarity reference setting unit that sets a reference for calculating a similarity based on a distribution of a texture feature amount of a learning window in a feature space.