JP5848539B2 - Ultrasonic data processor - Google Patents

Description

  The present invention relates to an ultrasonic data processing apparatus, and more particularly to calculation of the displacement of a point of interest between frames (or between volumes).

  An ultrasonic diagnostic apparatus is an apparatus that forms an ultrasonic image based on data obtained by transmitting and receiving ultrasonic waves to and from a living body. More specifically, an ultrasonic beam is scanned to sequentially form a beam scanning surface (two-dimensional capture space). As a result, a received frame sequence according to the transmission / reception coordinate system and the transmission / reception rate is acquired. Each received frame is composed of a plurality of beam data, and each beam data is composed of a plurality of echo data. The received frame sequence is converted into a display frame sequence by the digital scan converter. The display frame sequence follows the display coordinate system and the display frame rate. In the digital scan converter, a pixel group necessary to cope with a difference in data arrangement between a reception frame and a display frame is generated by interpolation processing. When the sector scanning method or convex scanning method in which the scanning surface extends in the depth direction is executed, the actual data density is lower in the deeper part than the shallow part, so more interpolation pixels are arranged in the deeper part than in the shallow part. Is done. The ultrasonic diagnostic apparatus is a concept including an ultrasonic data processing apparatus. The ultrasonic data processing apparatus may be configured as an information processing apparatus.

  Incidentally, an inter-frame matching process is known as a technique for detecting the movement of a point of interest in a time-series frame sequence (received frame sequence, display frame sequence). It calculates the displacement (two-dimensional displacement vector) of a point (attention point) between individual frames in the frame sequence (see Patent Document 1). As the technique, a block matching method (cross-correlation method, minimum sum absolute difference method, etc.) is well known. For example, the correlation index is calculated while changing the relative position between the two display frames, and the shift amount Δx in the horizontal direction when taking a peak in the two-dimensional distribution of the correlation index obtained thereby, the vertical direction The shift amount Δy is specified and used as the displacement (Δx, Δy). The peak referred to here is the maximum value of the correlation index in the cross-correlation method, and the minimum value of the correlation index in the minimum sum absolute difference method. Δx and Δy are each calculated as an integer value. The matching processing as described above can be applied to the received frame sequence in addition to the display frame sequence. In this case, the shift amount Δθ in the beam scanning direction (azimuth direction) and the shift in the beam depth direction between the received frames. The quantity Δr is calculated. In this case, Δθ and Δr are calculated as integer values.

  Since each display frame constituting the display frame sequence is after interpolation processing, that is, each pixel value constituting the display frame is basically an interpolation value and not an actual value. In processing, displacement calculation accuracy is likely to be a problem. Therefore, it is desirable to apply inter-frame matching processing to the received frame sequence to obtain the displacement between frames. However, in the received frame sequence obtained by sector scanning, convex scanning, etc., the beam data interval increases in deeper regions. When the matching process for obtaining the integer value is performed, there is a problem that the displacement estimation accuracy is considerably lowered particularly in the scanning direction.

  In order to cope with this problem, it is conceivable to apply a known sub-pixel process subsequent to the matching process. The sub-pixel process is a process for estimating a true peak, and is a process for obtaining a decimal with respect to an integer. According to this process, the displacement can be calculated as a real number obtained by adding a decimal number to an integer. However, it is desirable to improve the estimation accuracy where the beam interval such as a deep part is widened. From the viewpoint of processing efficiency, it is desirable to switch the estimation method according to the situation.

  Non-Patent Document 1 describes a highly accurate subpixel estimation method. However, it does not describe application to ultrasonic images or adaptive selection of a plurality of estimation methods.

JP 2006-325686 A

Shimizu et al., Precise subpixel estimation method for image matching, IEICE Transactions, d-II Vol. J84-D-II No.7, pp.1409-1418, 2001.7

  An object of the present invention is to increase the accuracy of displacement calculation particularly in a deep part when a beam scanning space having a wide end is formed and displacement calculation is performed on data before coordinate conversion. Alternatively, an object of the present invention is to apply a displacement estimation method suitable for the situation.

  The present invention relates to an ultrasonic data processing apparatus that calculates the displacement of a point of interest between a first beam data group and a second beam data group acquired by sequentially forming an ultrasonic beam scanning space that extends in the depth direction. In the above, based on the result of the first matching process between the first beam data group and the second beam data group, an integer value representing the displacement of the point of interest and an integer value in the depth direction and a scanning direction By applying the first sub-pixel process to the result of the first matching process and the integer value calculating means for calculating the integer value, at least a first small value in the scanning direction as a decimal part representing the displacement of the attention point. Based on a result of a second matching process between the first decimal value calculation means for calculating a numerical value and the first beam data group and the second beam data group, the displacement of the attention point Based on the first decimal value and the second decimal value, the second decimal value calculating means for calculating at least the second decimal value in the scanning direction as the decimal part to be represented, and the high precision decimal part representing the displacement of the point of interest A high-precision decimal value calculation means for calculating a high-precision decimal value, and an output unit capable of outputting real number information based on the integer part and the high-precision decimal part as the displacement of the attention point, .

  According to the above configuration, the displacement of the attention point is calculated between the first beam data group and the second beam data group. Each beam data group preferably constitutes a two-dimensional reception frame or three-dimensional volume data. When calculating the displacement, first, a first matching process is executed, whereby integer values in the depth direction and the scanning direction are calculated as integer parts. In this case, a pattern matching method is usually used, a template including a point of interest is set on one frame, and search blocks (blocks corresponding to the template) are sequentially set on the other frame while changing their positions. The pattern matching results between them are calculated sequentially. The integer part of the displacement of the point of interest is specified as the position that produces the best matching result. However, with only the integer part, the displacement estimation accuracy is greatly reduced where the beam interval is widened particularly in a deep part. Therefore, sub-pixel processing is applied to the calculation result of the integer part, and the decimal part added to the integer part is calculated. In this case, the decimal value is calculated at least in the scanning direction, but the decimal value may be further calculated in the depth direction. As described above, the displacement of the point of interest is estimated as a real number as a combination of the integer part and the decimal part.

  In sub-pixel processing, variation in error is recognized, and it is desired to estimate the displacement with higher accuracy by reducing it. Therefore, in the present invention, a second matching process different from the first matching process is applied between the first beam data group and the second beam data group, and a second decimal value is calculated by applying a sub-pixel process. Like to do. It has been confirmed that if the first decimal value and the second decimal value are taken into account together, the error in the decimal value calculation can be reduced, and using this property, the displacement of the attention point can be estimated with higher accuracy. To do. The calculation of the second decimal value is preferably performed at least in the scanning direction, but may be performed in the depth direction.

  Preferably, the first beam data group and the second beam data group constitute a first reception frame and a second reception frame, and coordinate conversion is performed with an interpolation process on the first reception frame and the second reception frame. It includes conversion means for executing processing to generate a first display frame and a second display frame, and the displacement calculation means is provided in a preceding stage of the conversion means. Since an error occurs due to the interpolation processing in the coordinate conversion processing, it is desirable to calculate the displacement before the coordinate conversion processing.

  Preferably, the high precision decimal value calculation means calculates the high precision decimal value as an average value of the first decimal value and the second decimal value. Therefore, when calculating the second decimal value, the calculation conditions are determined so that a high-precision decimal value can be obtained by comparison and integration with the first decimal value. Preferably, a half-pixel correction process is applied at that time, and the shift direction is determined by the first decimal value.

  Preferably, the second decimal value calculating means includes means for generating a corrected beam data group obtained by shifting the first beam data group by a half pixel, and between the corrected beam data group and the second beam data group. Means for calculating at least a second decimal value in the scanning direction by executing the second matching process as the same process as the first matching process. In the calculation of the second decimal value, a half-pixel addition / subtraction process corresponding to a half-pixel shift is applied.

  Preferably, the output unit selectively outputs real number information based on the integer part and the high-precision decimal part and real number information based on the integer part and the first decimal part as the displacement of the attention point. Can do. Preferably, the output unit may further selectively output the integer part as the displacement of the attention point. Preferably, the output unit switches the output target according to at least the depth of the attention point. Desirably, the output target is switched according to at least the beam interval. It is desirable to dynamically switch the calculation method in consideration of the calculation amount and the estimation accuracy.

  According to the present invention, when a divergent beam scanning space is formed, it is possible to enhance the calculation of the target point displacement estimation particularly in a deep part.

1 is a block diagram showing a preferred embodiment of an ultrasonic data processing apparatus (ultrasonic diagnostic apparatus) according to the present invention. It is a block diagram which shows the structural example of the displacement estimation part shown in FIG. It is a figure for demonstrating the matching process between frames. It is a graph of the SAD value which shows the result of a matching process. It is a figure for demonstrating the sub pixel process (parabolic fitting method) which calculates | requires a decimal value. It is a figure which shows the calculation range of a SAD value at the time of applying a one-dimensional parabolic fitting method to an azimuth | direction direction. It is a figure which shows the calculation range of a SAD value at the time of applying a one-dimensional parabolic fitting method to a depth direction. It is a figure which shows the error function by the 1st decimal estimation part, and the error function by the 2nd decimal estimation part on the assumption of a half pixel shift. It is a figure for demonstrating the production | generation of the 2nd template block accompanied by half pixel correction. It is a figure for demonstrating operation | movement of three estimation parts, and operation | movement of an output part.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described with reference to the drawings.

  FIG. 1 shows a preferred embodiment of an ultrasonic data processing apparatus according to the present invention. This ultrasonic data processing apparatus is configured as an ultrasonic diagnostic apparatus.

  The probe 10 is used in contact with the body surface of a subject and transmits and receives ultrasonic waves. The probe 10 includes a 1D array transducer. The 1D array transducer is composed of a plurality of vibrating elements, thereby forming an ultrasonic beam, which is electronically scanned. In this embodiment, electronic sector scanning is applied. In addition, convex scanning or the like may be applied. In these scanning methods, a plurality of beam lines extending in a fan shape are formed, and the beam interval increases as the part is deeper. In FIG. 1, the sector scanning plane is shown, r indicates the depth direction, and θ indicates the electronic scanning direction. The scanning plane is repeatedly formed, and a reception frame (arrangement of a plurality of beam data) is formed each time.

  The transmission / reception unit 12 functions as a transmission beam former and a reception beam former. At the time of transmission, a plurality of transmission signals are supplied to the array transducer, thereby forming a transmission beam. At the time of reception, the reflected wave from the living body is received by the array transducer, and a plurality of reception signals are output from the array transducer. The transmission / reception unit 12 generates beam data by phasing addition processing for a plurality of reception signals. The beam data is output to the image forming unit 18 side.

  The displacement estimation unit 14 is a module that calculates the displacement (two-dimensional movement vector) of the target point between adjacent frames, and the function is realized by, for example, a software function. Although the specific contents of the displacement estimation unit 14 will be described in detail later, information on the estimated displacement is output to the displacement correction unit 16, the image forming unit 18, and the display processing unit 20 as necessary. The displacement correction unit 16 applies correction processing to the frame using the estimated displacement. When each pixel is set as the target pixel, the movement of each target pixel can be canceled by correction using the displacement of each target pixel. The displacement correction unit 16 is provided as necessary.

  The image forming unit 18 includes a digital scan converter (DSC), applies coordinate conversion processing and interpolation processing to each received frame, and sequentially generates display frames. The image forming unit 18 forms a B-mode image. The display frame is sent to the display processing unit 20. When frame correlation processing is performed in the image forming unit 18, the frame correlation processing may be performed while canceling the displacement by using the information on the displacement. The display processing unit 20 has an image composition function, and also has a function of forming a graph and other images based on the displacement. An ultrasonic image is displayed on the display unit 22. The control unit 24 includes a CPU and an operation program, and controls the operation of each component shown in FIG. The input unit 26 is configured by an operation panel.

  FIG. 2 shows a configuration example of the displacement estimation unit 14 shown in FIG. The displacement estimation unit 14 includes a block extraction unit 30, a matching execution unit 32, an integer estimation unit 34, a first decimal number estimation unit 36, a second decimal number estimation unit 38, an output unit 40, a size determination unit 42, and an operation control unit 44. Have. The size determination unit 42 adaptively sets the size of the template and search block to be referred to in the matching process based on the information of the line density D1. However, the size may be made constant regardless of the line density or the depth direction. The motion control unit 44 switches the displacement estimation method based on the information on the block depth R1 and the information on the line density D1, and specifically, the first decimal number estimation unit 36 and the second decimal number estimation unit 38. The operation is controlled on and off. The output unit 40 also refers to the presence or absence of operation. It is desirable to increase the estimation accuracy step by step depending on the depth. If a simple estimation method is selected in a shallow region, it is efficient because the amount of calculation can be reduced. The block extraction unit 30, matching execution unit 32, and integer estimation unit 34 shown in FIG. 2 execute an integer value calculation method described below. The first decimal fraction estimation unit 36 executes a first subpixel estimation method described below. The second decimal number estimation unit 38 executes a second subpixel estimation method described below. It actually has functions corresponding to the block extraction unit 30 and the matching execution unit 32, but these are not shown in FIG. Pre-displacement data and post-displacement data are also input to the second decimal number estimation unit 38 (in the conceptual diagram shown in FIG. 1, these data are transmitted via the integer estimation unit 34 and the first decimal number estimation unit 36. Sent to the second decimal number estimation unit 38).

(1) Integer value calculation method: In the block matching (Block-Matching) method, a first template block R (θ, r) including a point of interest is set on a reference image before displacement (actually a received frame), A brute force search is performed as to where in the search area the first template block R (θ, r) has moved on the image after displacement (actually a received frame). FIG. 3A shows a part of the received frame before displacement. FIG. 3B shows a part (search area) of the received frame after displacement. In the figure, each grid point represents a data coordinate. On the received frame before displacement, a template block 46 having a predetermined shape with the attention point as the center is set. On the other hand, in a subsequent received frame, a large search area is set around the corresponding point corresponding to the point of interest, and the search block 48 is sequentially moved by one pixel in each direction within the range (for example, raster scan is executed). The matching process is executed. Thereby, the correlation index is calculated at each position. As correlation indexes, cross-correlation and the sum of absolute differences (SAD) are widely known. Here, the SAD shown in the following equation (1) will be described. The smaller the SAD value, the higher the correlation. Therefore, as shown in FIG. 4, a search block having a minimum value of SAD ₁ (Δθ, Δr) is set as a matching block (search block at the optimal position), and the coordinate difference between the template block and the matching block is expressed in integer units. The estimated displacement (Δθ ₀ , Δr ₀ ) is assumed (FIG. 4 shows the SAD function for the azimuth direction). When cross-correlation is adopted, the higher the value is, the higher the correlation is evaluated. Therefore, the search block having the maximum value is set as the matching block. In the following description, (θ, r) is the coordinate of each block center point (V0 (θ, r) & V1 (θ, r)).

In (1) above, the block size = MB * MB, the search area size = MA * MA, the SAD ₁ size = MC * NC = (MA-MB + 1) * (NA-NB + 1), The range of Δθ = −MC / 2 to MC / 2−1, and the range of Δr = −NC / 2 to NC / 2−1. (Δθ ₀ , Δr ₀ ) = SAD ₁ (Δθ, Δr).

In FIG. 2, the block extraction unit 30 sets a template block for the received frame V0 (θ, r) before displacement, and the search area S (θ, r) for the received frame V1 (θ, r) after displacement. r) and a search block are set. The matching execution unit 32 calculates SAD as a correlation index between the template block and the search block set at each position. The integer estimation unit 34 obtains an estimated displacement (Δθ ₀ , Δr ₀ ) as an integer part based on SAD ₁ (Δθ, Δr) having the best value.

(2) First sub-pixel estimation method: From SAD ₁ (Δθ, Δr) obtained by the above-described integer value calculation method and estimated displacement (Δθ ₀ , Δr ₀ ) in integer units, a sub-pixel as a decimal part of displacement (Sub-Pixel (less than an integer)) (dθ ₁ , dr ₁ ) is estimated. Note that (dθ ₁ , dr ₁ ) is only a component less than an integer that does not include an integer component. That is, | dθ ₁ | ≦ 0.5 and | dr ₁ | ≦ 0.5.

As the subpixel estimation method, many methods such as a cubic spline interpolation method and a parabolic fitting method have been proposed. Here, the most common parafitting method will be described. In this method, assuming that the SAD distribution obtained by the integer value calculation method is a quadratic curve, as shown in FIG. 5, the sub-pixel is calculated by the following equations (2) and (3). It is. Note that the number of SAD ₁ (Δθ, Δr) used in the parabolic fitting method for each direction is only three before and after the displacement (Δθ ₀ , Δr ₀ ) in integer units. They are illustrated in FIGS.

The first decimal number estimation unit shown in FIG. 2 performs the above calculation. As a result of estimation, a decimal part (dθ ₁ , dr ₁ ) of the displacement is obtained.

(3) Second sub-pixel estimation method (high-precision sub-pixel estimation method, half-pixel correction method): In any of the above-mentioned traditional sub-pixel estimation methods, an error (Pixel) biased to an integer unit or ± 1/2 -Called Locking phenomenon). In this regard, as shown in FIG. 8, the estimation error in the first template block and the sub-pixel estimation error due to the second template block g (m, n) translated 0.5 pixels in the positive and negative directions are almost positive and negative symmetric. The half-pixel correction uses this characteristic, and calculates the sub-pixel estimation value dθ ₁ by the first template block and the sub-pixel estimation value dθ ₂ by the second template block by applying the integer value calculation method and the sub-pixel estimation method. Average to reduce the sub-pixel estimation error. The procedure is as follows.

First, a second template block g (m, n) translated by ± 0.5 is created. This is shown in FIG. If the first decimal part d [theta] ₁ is negative, the second template block is +0.5 translate into positive generating by linear interpolation (in FIG. 9 (A)). If the first decimal part d [theta] ₁ is positive, a second template block -0.5 is translated to the negative side to generate a linear interpolation process (of FIG. 9 (B)). FIG. 9C shows the received frame after displacement. If the above processing is expressed by a mathematical formula, the following formula (4) is obtained.

Next, a correlation index is calculated using the second template block g (m, n). In that case, the same correlation index as that used in the integer value calculation method is used. Since SAD is used in the above example, SAD is calculated here. That is, the following equation (5) is calculated. Incidentally, from the viewpoint of reducing the amount of calculation, it is desirable to use Δr ₀ obtained by the integer value calculation method as it is in the distance direction (depth direction) and perform a search only in the vicinity of the azimuth direction Δθ ₀ . Since only three SAD ₂ (Δθ) by the parabola fitting method are required, Δθ = [− 1, 0, +1], an increase in processing time to apply the half-pixel correction method is not so large.

Then, to calculate the sub-pixel d [theta] ₂ of the second template block g (m, n). In this case, the same estimation method as the above subpixel estimation method is used. That is, in this example, the parabolic fitting method is used, and the following equation (6) is calculated.

Next, as shown in the following equation (7), the ± 0.5 translation is subtracted to obtain | dθ ₂ | ≦ 0.5. At this stage, the error between the d [theta] ₁ and d [theta] ₂ is approximately positive-negative symmetrical (see Figure 8).

  The second decimal number estimation unit 38 shown in FIG. 2 executes the above equations (4) to (7).

  The output unit 40 shown in FIG. 2 synthesizes or selects outputs according to the conditions shown in FIG. That is, in the first case, only the integer estimator functions, and the integer part estimated thereby is the final output. For example, this condition is applied to a shallow region and a region with a high beam density. In the second case, the integer estimator and the first fraction estimator function, and the integer part and the first fraction part estimated thereby are combined and output. For example, this condition is applied to an intermediate region and a region with an intermediate beam density. In FIG. 10, decimal values are added to the integer values in both the azimuth direction and the depth direction. However, arithmetic and addition of decimal values may be performed only in one direction. In the third case, the integer estimation unit, the first decimal number estimation unit, and the second decimal number estimation unit function. In particular, an average value of the first decimal value and the second decimal value is calculated, which can be called a high-precision decimal value. A high-precision decimal value is added to the integer value, which is output. For example, this condition is applied to a deep region and a region with a low beam density. In the example shown in FIG. 10, high-precision decimal values are calculated for the azimuth direction, but such calculation may be performed for both directions (in FIG. 2, the second decimal value is calculated for both directions). ).

  Although the above-described embodiment is processing for two-dimensional data, the same processing as described above can be applied to three-dimensional data.

  14 displacement estimation part, 34 integer estimation part, 36 1st decimal estimation part, 38 2nd decimal estimation part, 40 output part.

Claims (5)

In an ultrasonic data processing apparatus for calculating a displacement of a point of interest between a first beam data group and a second beam data group acquired by sequentially forming an ultrasonic beam scanning space extending in a depth direction,
Based on the result of the first matching process between the first beam data group and the second beam data group, an integer value in the depth direction and an integer value in the scanning direction are used as integer parts representing the displacement of the point of interest. Integer value calculating means for calculating
First decimal value calculation means for calculating at least a first decimal value in the scanning direction as a decimal part representing the displacement of the target point by applying a first sub-pixel process to the result of the first matching process; ,
Based on the result of the second matching process between the first beam data group and the second beam data group, at least a second decimal value in the scanning direction is calculated as a decimal part representing the displacement of the point of interest. 2 decimal value calculation means;
High-precision decimal value calculation means for calculating a high-precision decimal value based on the first decimal value and the second decimal value as the high-precision decimal part representing the displacement of the point of interest;
An output unit capable of outputting real number information based on the integer part and the high-precision decimal part as the displacement of the attention point;
Only including,
The output unit can selectively output real number information based on the integer part and the high-precision decimal part and real number information based on the integer part and the first decimal value as the displacement of the attention point. And
The output unit switches the output target according to at least the depth of the attention point or the beam interval,
An ultrasonic data processing apparatus. The apparatus of claim 1.
The first beam data group and the second beam data group constitute a first reception frame and a second reception frame;
Conversion means for generating a first display frame and a second display frame by performing a coordinate conversion process with an interpolation process on the first reception frame and the second reception frame;
Means for calculating the displacement is provided in front of the converter, ultrasound data processing apparatus according to claim. The apparatus according to claim 1 or 2,
The ultrasonic data processing apparatus, wherein the high-precision decimal value calculation means calculates the high-precision decimal value as an average value of the first decimal value and the second decimal value. The device according to any one of claims 1 to 3,
The second decimal value calculation means includes:
Means for generating a corrected beam data group obtained by shifting the first beam data group by a half pixel;
Means for calculating at least a second decimal value in the scanning direction by executing the second matching process as the same process as the first matching process between the correction beam data group and the second beam data group. When,
An ultrasonic data processing apparatus comprising: The apparatus of claim 1 .
The ultrasonic data processing apparatus, wherein the output unit can further selectively output the integer part as a displacement of the attention point.