JP2005165627A - Ultrasonic diagnostic apparatus, and volume data processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To raise data transfer efficiency without deteriorating volume rendering calculation when data is transferred from an external memory to a 3D processor so as to perform the volume rendering calculation in an ultrasonic diagnostic apparatus. <P>SOLUTION: Volume data in a 3D data space is stored in a 3D memory 24, and transferred to an internal memory 28 with slice data as unit. The 3D processor 26 determines a ray to perform voxel calculation at each data transfer, and advances the voxel calculation concerning the ray to a voxel calculation performable extent. A slice data taking-out direction is selected in response to the direction of the ray. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an ultrasonic diagnostic apparatus and a volume data processing method, and more particularly to a technique for forming a three-dimensional image based on volume data stored in a three-dimensional memory.

  When forming a three-dimensional ultrasonic image in the ultrasonic diagnostic apparatus, ultrasonic waves are transmitted / received to / from a three-dimensional transmission / reception space in the living body, thereby acquiring volume data. The volume data is composed of a plurality of voxel data (echo data). When forming a three-dimensional ultrasonic image using such volume data, for example, a volume rendering method is used.

  In this method, generally, a viewpoint and a screen are defined so as to sandwich a three-dimensional data space (a data space in which volume data exists), and a plurality of rays (lines of sight) are defined from the viewpoint to the screen. Then, for each ray, the voxel data existing on that ray is sequentially read out from the three-dimensional data memory, and voxel operation (here, output using the opacity based on the volume rendering method) is performed on each voxel data. Light amount calculation) is executed sequentially. A final voxel calculation result (output light amount) is converted into a pixel value, and a pixel value for each ray is mapped on the screen to form a three-dimensional ultrasonic image.

  Conventionally, dedicated hardware (a high-speed computing unit with a fixed function) is designed for various computations for forming a three-dimensional image, and the voxel computation and the like are executed by the hardware.

  On the other hand, recently, the processing speed of programmable general-purpose processors (for example, DSP (digital signal processor)) has improved, and functional design and changes are easy. It is requested to do.

  When realizing the above three-dimensional image processing with a general-purpose processor, generally, volume data is stored in a three-dimensional data memory (external memory), the general-purpose processor accesses the three-dimensional data memory, and voxels are set for each ray. It is necessary to specify the voxel data necessary for the calculation each time and import it into the general-purpose processor. If the viewpoint can be set at an arbitrary position, each ray will pass through the volume data diagonally in most cases, but in that case, if voxel calculation is performed simultaneously for each ray, voxels for multiple rays are used. A group of voxel data necessary for the calculation is variously distributed on the volume data, that is, it is necessary to perform random access to obtain the necessary voxel data for the three-dimensional data memory.

  In addition, when a memory device such as SRAM is used as a three-dimensional data memory, such random access can be performed at high speed. Generally, however, a low-cost and highly integrated DRAM or SDRAM is used. If random access is attempted in the case of the above memory device, the read time increases, resulting in a result contrary to the request for high-speed transfer.

  Patent Document 1 below discloses a technique for forming a three-dimensional ultrasonic image based on a volume rendering method in an ultrasonic diagnostic apparatus. However, there is no specific disclosure regarding the data transfer method and how the voxel calculation proceeds for each ray. In Patent Document 2 below, data transfer from a three-dimensional data space (main memory) to an image processor is performed in units of data blocks having a specific address relationship in order to prevent the occurrence of a cache miss. An apparatus is disclosed. In this apparatus, a special calculation is performed at the time of reading data to obtain a set of read addresses, and the necessity of random access can be pointed out.

Japanese Patent No. 2883584 JP-A-9-134447

  An object of the present invention is to enable efficient formation of a three-dimensional ultrasonic image based on volume data.

  Another object of the present invention is to enable easy and quick data transfer from a three-dimensional data memory to a processor for image processing.

(1) The present invention relates to a transmission / reception means for transmitting and receiving ultrasonic waves to and from a living body, thereby capturing volume data composed of a plurality of voxel data, a three-dimensional data memory storing the volume data, Three-dimensional image forming means for forming a three-dimensional image based on the volume data, and means for transferring the volume data from the three-dimensional data memory to the three-dimensional image forming means, wherein a predetermined data set is transferred in units Data transfer means for repeatedly executing data transfer as, the three-dimensional image forming means, a ray setting means for setting a plurality of rays in the three-dimensional data space, and for each data transfer, the plurality of data transfer means Ray determining means for determining a ray capable of executing voxel calculation in the rays of the data, and executing the voxel calculation for each data transfer Voxel calculation means for advancing the voxel calculation to the coordinates at which voxel calculation can be performed on a, and the three-dimensional image is formed as a result of the last voxel calculation for each ray, and voxel calculation is independent for each ray And proceed.

  According to the above configuration, data transfer from the three-dimensional data memory to the image forming unit is performed in units of a predetermined data set, and a voxel operation that can be executed for each data transfer is performed. That is, for each data transfer, a ray that can execute the voxel operation is determined among the plurality of rays, and the voxel operation is performed up to the coordinates that can execute the voxel operation for the ray. Thereby, it is possible to realize extremely efficient arithmetic processing according to the contents of the transferred data.

  For example, if you want to execute voxel operations with the same degree of progression for each ray, or specify the necessary data for each ray and acquire it sequentially to perform voxel operations, it is inevitably complicated address Calculation is required. On the other hand, according to the present invention, since it is only necessary to execute a voxel operation that can be executed depending on the transferred data set, the data transfer is simplified, that is, a complicated address operation is sequentially performed. There is no need.

  The wave transmitting / receiving means includes a wave transmitter / receiver that uses both electronic scanning and mechanical scanning, a wave transmitter / receiver that performs two-dimensional electronic scanning, and the like. The three-dimensional data memory is preferably composed of a memory suitable for block transfer of data. The plurality of rays may be set parallel to each other or may be set non-parallel. The predetermined data set is preferably a data block in which data is continuous in the memory. However, as long as an operation according to the above method can be executed for each data transfer to finally form a three-dimensional image, one or a plurality of data sets are stored. It may have a continuous portion. Here, it is desirable to use one or a plurality of slice data as a transfer unit, and in particular, it is preferable to use one slice data as a transfer unit. However, continuous reading is not performed in the three-dimensional data space or the memory address space. As far as possible, for example, 1.5 slice data can be used as a transfer unit. The above three-dimensional image forming means is preferably constituted by a program operation type processor. In this case, the data transfer means may be realized as a function of the processor.

  Preferably, the predetermined data set is slice data composed of a plurality of voxel data having continuity in a three-dimensional data space, and the data transfer means repeats data transfer in the arrangement order of the slice data. Execute. The slice data is, for example, each surface data or block data arranged on a specific axis in a three-dimensional coordinate axis.

  Preferably, the data transfer means includes selection means for selecting a positive direction or a negative direction in the arrangement direction of the slice data as the extraction direction of the slice data. Preferably, the direction selection means selects the forward direction or the reverse direction according to the position of the viewpoint with respect to the three-dimensional data space.

  Since the voxel calculation proceeds in the direction of the line of sight on the ray (line of sight), the direction along the line of sight among the positive direction and the negative direction in the arrangement direction is selected as the extraction direction. Exceptionally, the ray and the alignment direction are orthogonal to each other. In that case, the voxel calculation can be performed in any direction. In this case, basically, when one data set is obtained, all voxel operations are completed only for each ray passing through the data set.

  Preferably, the data storage space includes coordinate conversion means for performing coordinate conversion from the coordinate system of the three-dimensional transmission / reception space to the coordinate system of the three-dimensional data space when the volume data is written to the three-dimensional data memory. Has three orthogonal coordinate axes, and the slice data corresponds to a plane orthogonal to a specific coordinate axis among the three orthogonal coordinate axes.

  Coordinate conversion can be performed at the time of writing volume data and at the time of reading volume data. However, if coordinate conversion is performed at the time of writing, it becomes easy to block transfer a continuous data set on the memory.

  Preferably, the three-dimensional image forming means includes data interpolation means for generating data necessary for the voxel calculation by interpolation processing. When voxel data does not exist on the coordinates to be calculated on the ray, interpolated data is generated from two or more existing data by two-dimensional or three-dimensional interpolation and used in voxel calculation It is. Instead of the interpolation calculation, it is possible to use a method of using the nearest data as it is and using the same data a plurality of times as necessary. When performing three-dimensional interpolation, a plurality of data sets may be used as necessary. That is, for example, necessary interpolation data may be generated using the data set transferred last time and the data set transferred this time. Even in such a case, if the data set is configured as a data string having continuity in the three-dimensional data space or the memory address space, the transfer of each data set can be easily performed.

  Preferably, the three-dimensional image forming means includes means for equalizing the number of voxel operations per unit length for each ray. According to this configuration, it is possible to make the calculation pitch for each ray uniform and to stabilize the image quality regardless of the direction of the ray. That is, it is possible to prevent the problem that the eye feel and the feeling of depth differ depending on the position of the viewpoint.

  Preferably, the three-dimensional image forming means includes a plurality of three-dimensional image forming modules operating in parallel, wherein the plurality of rays are divided into a plurality of groups, and each of the three-dimensional image forming modules is associated with the group. In charge of voxel operations. In this case, a plurality of rays may be allocated to a plurality of groups in units of one ray, a plurality of rays may be allocated to a plurality of groups in units of one ray row, or one ray block A plurality of rays may be allocated to a plurality of groups in units of.

(2) Further, the present invention provides a transmission / reception means for transmitting / receiving ultrasonic waves to / from a three-dimensional transmission / reception space of a living body, thereby acquiring volume data composed of a plurality of voxel data, and three coordinate axes orthogonal to each other. A three-dimensional data space having the volume data stored therein, three-dimensional image forming means for forming a three-dimensional image based on the volume data, and the third order from the three-dimensional data memory. Means for transferring the volume data to the original image forming means, wherein slice data orthogonal to specific coordinate axes of the three coordinate axes in the three-dimensional data space is used as a transfer unit, and the direction of the specific coordinate axes is Data transfer means for repeatedly executing data transfer as a slice data take-out direction, A forming unit configured to set a plurality of rays for the three-dimensional data space; an internal memory for temporarily storing the slice data transferred; and an internal memory for each data transfer. The slice data stored in is used to advance the voxel calculation to the coordinates where the voxel calculation can be performed among the plurality of rays to the coordinates at which the voxel calculation can be executed, and as a result, the result of the last voxel calculation for each ray Means for obtaining pixel values constituting the three-dimensional image.

  According to the above configuration, since data transfer is performed in units of slice data, a complicated address calculation is not required to specify it. Here, the voxel calculation proceeds for each ray in units of slice data.

  Preferably, the three-dimensional image forming means further includes initial coordinates on a first axis in the three-dimensional data space as a point at which each ray passes first with respect to the three-dimensional data space, An initial coordinate calculating means for calculating an initial coordinate on the second axis and an initial coordinate on the third axis; and a step component of the first axis, a step component of the second axis, and a step component of the third axis. Component calculation means, and for each of the rays, step components on the first axis with respect to the initial coordinates on the first axis, the initial coordinates on the second axis, and the initial coordinates on the third axis, respectively. , By adding cumulatively the step component on the second axis and the step component on the third axis, target coordinate specifying means for specifying the voxel calculation target coordinate on each ray, and on the internal memory From the stored slice data, Extracts voxel data corresponding to the cell operation target coordinates, or includes, means for generating by interpolation voxel data corresponding to the voxel calculation target coordinates using the slice data stored on the internal memory.

  According to the above configuration, initial coordinates are obtained for each coordinate axis for each ray, and target coordinates on each ray are sequentially determined based on the initial coordinates.

  Preferably, a unit for obtaining a coordinate component on the specific coordinate axis for a unit ray vector defined for the three-dimensional data space, and when the coordinate component on the specific coordinate axis is positive, Selecting a positive direction as a slice data extraction direction, and selecting a negative direction of the specific coordinate axis as a slice data extraction direction when the coordinate component of the specific coordinate axis is negative.

  According to the above configuration, it is possible to automatically determine the extraction direction of slice data based on the unit ray vector. A single unit ray vector may be set for a plurality of rays, or a unit ray vector may be set for each of a plurality of rays.

(3) The present invention also provides a transfer step of transferring the volume data from a three-dimensional data memory storing volume data composed of a plurality of voxel data to an internal memory in the image processor, and the transferred volume An image forming step of forming a three-dimensional image based on a volume rendering method using the data, wherein the transfer step includes a plurality of slice data in which the volume data is aligned in a specific coordinate axis direction. And the step of transferring slice data in order in the direction of the predetermined coordinate axis from among the plurality of slice data, wherein the image forming step sets a plurality of rays for the three-dimensional data space For each step and each data transfer, a ray that can execute voxel computation among the plurality of rays. There allowed to proceed voxel calculation until the voxel calculation executable coordinates, thereby comprising a step of determining a pixel value forming the three-dimensional image as a result of the last voxel calculation for each ray, a.

  The above method is used particularly in the case of constructing a three-dimensional image based on the volume rendering method in an ultrasonic diagnostic apparatus, but can also be used in other fields that handle volume data.

  As described above, according to the present invention, it is possible to efficiently form a three-dimensional ultrasonic image based on volume data. According to the present invention, data transfer from a three-dimensional data memory to a processor for image processing can be performed easily and quickly.

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

  FIG. 1 shows the principle of the volume rendering method. The three-dimensional data space 36 has volume data acquired by two-dimensional scanning with an ultrasonic beam, and is virtually constructed on a 3D memory described later on the ultrasonic diagnostic apparatus. Here, the three-dimensional data space 36 has coordinate axes of x, y, and z that are orthogonal to each other, and voxel data (echo data) exists at each coordinate in the three-dimensional data space 36. When the three-dimensional ultrasonic transmission / reception space is configured as a polar coordinate space of r, θ, φ, the coordinates of (r, θ, φ) that each voxel data has (x, converted to the coordinates of y, z).

  In volume rendering, a viewpoint is normally set outside the three-dimensional data space 36, and a screen 40 as a two-dimensional plane is virtually set on the opposite side of the viewpoint through the three-dimensional data space. Is done. A plurality of rays (line of sight) are defined on the basis of the viewpoint (one ray 38 is shown as a representative in FIG. 1). The ray 38 penetrates the three-dimensional data space 36, and therefore, the ray 38 corresponds to a voxel data string composed of a plurality of voxel data. When the voxel calculation based on the volume rendering method is sequentially executed for each echo data from the viewpoint side, the pixel value is determined as a result of the final voxel calculation. The pixel value is mapped to a coordinate P corresponding to the ray 38 on the screen.

  The screen has X and Y coordinate axes, ie, each coordinate is defined by X and Y coordinates. Rays are set for each coordinate. As described above, if the pixel values obtained for each ray are mapped on the screen 40, a three-dimensional image is formed on the screen. The plurality of rays may be parallel to each other, but the plurality of rays may be non-parallel to each other.

In the volume rendering method, various types of calculation formulas for voxel calculation for each echo data are known. Basically, in any calculation formula, opacity is used as a parameter for each voxel calculation of each echo data. Using such parameters, an output light amount (output value) is obtained for each voxel operation, and this is used as an input light amount (input value) in the next voxel operation. This is then repeated. The output light amount at the time when the calculation end condition is satisfied is converted into a pixel value. That is, it is based on a model in which light propagates through a medium while being scattered and attenuated. In the present embodiment, for example, the following equation (1) is used.

  Of course, a formula other than the formula (1) may be used. For example, an arithmetic formula as described in Patent Document 1 may be used. The voxel calculation for each ray is performed, for example, when the target coordinate exceeds the three-dimensional data space, when the output light amount exceeds a predetermined value (for example, 1), and cumulative addition of the opacity used in each voxel calculation. The process ends when a predetermined condition is satisfied, such as when the value exceeds a predetermined value (for example, 1). The output light amount at that time is associated with the pixel value.

  Note that the data transfer and processing method according to the present embodiment is particularly suitable when performing image processing by the volume rendering method as described above, but can also be applied to an image processing method that performs sequential calculation along each ray. It is. For example, an integration method, a maximum value or minimum value detection method, and the like.

  FIG. 2 is a block diagram showing the overall configuration of the ultrasonic diagnostic apparatus according to this embodiment.

  The 3D probe 10 is an ultrasonic probe for taking in three-dimensional echo data. The 3D probe 10 is used in contact with the body surface or inserted into a body cavity. An ultrasonic beam is formed by the 3D probe 10, and the ultrasonic beam is two-dimensionally scanned. A three-dimensional transmission / reception space 12 is formed by two-dimensional scanning of the ultrasonic beam.

  Here, the transmission / reception space 12 is a three-dimensional space defined by three coordinates of r, θ, and φ. For example, when the ultrasonic beam B is scanned in the θ direction, a scanning surface S is formed. When scanning in the φ direction, a three-dimensional transmission / reception space 12 is formed. The 3D probe 10 may be a combination of electronic scanning and mechanical scanning, or may be one that two-dimensionally electronically scans an ultrasonic beam. In the latter case, a known 2D array transducer is used. In FIG. 2, the transmission / reception space 12 having a pyramid shape is shown, but the shape of the transmission / reception space 12 may be a cube or may be other than that.

  The transmission unit 14 functions as a transmission beam former. That is, the transmission unit 14 supplies a plurality of transmission signals to a plurality of vibration elements constituting the array transducer. The receiving unit 16 functions as a reception beam former, and executes phasing addition processing on a plurality of reception signals from a plurality of vibration elements constituting the array transducer. As a result, a reception signal is output from the reception unit 16 for each ultrasonic beam. The receiving unit 16 includes an A / D converter, and the received signal after phasing addition output from the receiving unit 16 is digital data. Here, each data corresponds to voxel data (echo data).

  The 2D conversion unit 18 is a module that forms a two-dimensional tomographic image (B-mode image) corresponding to the scanning plane S. The 2D conversion unit 18 is constituted by, for example, a conventionally used digital scan converter. The 2D conversion unit 18 executes processing for converting r, θ coordinates into x, y coordinates, other interpolation processing, and the like. The data of the two-dimensional tomographic image formed by the 2D conversion unit 18 is output to the display processing unit 20.

  The 3D conversion unit 22 performs three-dimensional coordinate conversion on the echo data output from the reception unit 16, that is, voxel data. That is, each voxel data has coordinates defined by r, θ, and φ, and the coordinates are converted into x, y, and z coordinates. That is, conversion from the polar coordinate space to the orthogonal coordinate space is performed. The voxel data after coordinate conversion is stored on the 3D memory 24.

  The 3D memory 24 has a memory space corresponding to the three-dimensional data space 36 shown in FIG. 1, and each address of the 3D memory 24 is associated with each coordinate in the three-dimensional space. Each input voxel data is stored at an address associated with the three-dimensional coordinates of the voxel data. The 3D memory 24 corresponds to an external memory from the viewpoint of the relationship with the 3D processor 26 described later. The 3D memory is configured by, for example, a DRAM, and is preferably configured by an SDRAM or the like. That is, it is desirable to use a memory capable of block transfer of stored data in predetermined data units as will be described later. In this case, the DMA transfer method can be used as the block transfer method.

  In the present embodiment, the 3D processor 26 is configured by a CPU or DSP (digital signal processor) that performs a program operation, and an internal memory 28 is provided therein. The internal memory 28 functions as a cache memory in the 3D processor 26, and the 3D processor 26 can access the internal memory 28 at high speed. On the other hand, data transfer from the 3D memory 24 to the internal memory 28 cannot generally be performed at a high speed.

  The 3D processor 26 incorporates a DMA controller in this embodiment, and the DMA controller can perform DMA transfer of data from the 3D memory 24 to the internal memory 28. As indicated by reference numeral 25, the 3D processor 26 may control the writing of data to the 3D memory 24. In this case, each data output from the 3D conversion unit 22 is once taken into the 3D processor 26. After that, it is stored in the 3D memory 24 via it. Of course, such a data writing to the 3D memory 24 may be performed by another processor, or may be realized as a function of the control unit 32 shown in FIG.

  In this embodiment, as will be described in detail later, data transfer from the 3D memory 24 to the internal memory 28 is executed in units of a predetermined data set, specifically slice data. The positive direction or the negative direction is appropriately selected from the arrangement directions of the slice data. In the 3D processor 26, a voxel operation that can be executed using the transmitted data is sequentially executed instead of performing the voxel operation for each ray at the same time as in the prior art. Therefore, although depending on the cross relationship between the three-dimensional data space and the ray, the degree of progress of the voxel operation is usually different for each ray.

  By using such a method, the transfer unit from the 3D memory 24 to the internal memory 28 can be set uniformly, and voxel calculation is advanced for each ray using the data obtained thereon. be able to. Since the 3D processor 26 can access the internal memory 28 at a high speed, random access to the internal memory 28 is not a burden on the operation, but random access from the 3D processor 26 to the 3D memory 24 is data. It becomes a bottleneck in image processing due to a low transfer rate. On the other hand, according to the method of the present embodiment, it is possible to efficiently perform image processing while avoiding such a problem by performing block transfer of data from the 3D memory 24 to the internal memory 28. It becomes. This will be described in detail later.

  In the present embodiment, the 3D processor 26 uses, for example, the expression shown in the above expression (1) as an arithmetic expression for voxel arithmetic. Therefore, the arithmetic expression is sequentially executed for each ray, and the output value when the end condition is satisfied is converted into a pixel value. The pixel value is stored in the frame memory provided in the display processing unit 20 from the 3D processor 26. Of course, such a frame memory may be provided in the 3D processor 26. In any case, a three-dimensional image is formed on the screen by repeatedly executing the voxel operation for each ray.

  The display processing unit 20 has a function for synthesizing image data and graphic data and other display processing functions. Depending on the operation mode of the apparatus, the display processing unit 20 can display 2D tomographic image data or 3D image data. Output data. The display unit 32 displays both or one of the two-dimensional tomographic image and the three-dimensional image.

  The control unit 32 is constituted by a CPU and an operation program therefor, and performs operation control of each component shown in FIG. An input unit 34 constituted by an operation panel is connected to the control unit 32. The user can use the input unit 34 to perform various input operations such as mode selection and parameter specification. Incidentally, the configuration for performing Doppler processing is omitted in FIG.

  FIG. 3 shows a 3D data space 36 virtually constructed on the 3D memory 24. The 3D data space 36 has an x-axis, a y-axis, and a z-axis, and each voxel data is stored in three-dimensional coordinates associated therewith. Here, addresses in the x-axis direction are defined as 0 to 1, addresses in the y direction are defined as 0 to m, and addresses 0 to n in the z direction are defined.

  FIG. 4 shows the address structure 38 of the 3D memory 24 shown in FIG. 1, and each voxel data is continuously stored on the one-dimensional address structure 38 as shown. Here, D corresponds to slice data, that is, data corresponding to one yz plane defined by a specific address on the x coordinate. Data transfer from the 3D memory 24 to the internal memory 28 is performed in units of such blocks. Of course, a plurality of slice planes may be used as one block, and transfer may be performed as a unit, or 1.5 slice planes may be used as one transfer unit. In any case, by transferring block data as it is, it is possible to quickly and easily block transfer the target data set only by specifying the start address and transfer capacity (or end address). . Therefore, the data transfer rate can be increased from several times to several tens of times as compared with the conventional random access method.

  However, simply performing block transfer makes it difficult to perform volume rendering operations for each ray in the 3D processor 26. That is, depending on the case, there is a problem that the progress of such calculation is obstructed. Therefore, in the present embodiment, a method of selecting the slice data extraction direction according to the ray direction is employed.

  FIG. 5 shows a ray unit vector V. This ray unit vector V is used for determining the direction of the ray and sequentially obtaining the coordinates to be calculated in each ray. This unit ray vector is set in common when a plurality of rays are set in parallel, and is set for each ray when the plurality of rays are set non-parallel. Of course, even in that case, one unit ray vector may be set as a representative.

  For example, for a representative ray, this unit ray vector V is defined as a vector having the same unit direction and unit size of 1. The components on the coordinate axes of the unit vector are Δx, Δy, and Δz.

  FIG. 6 shows a state in which a plurality of rays are set for the three-dimensional data space 36. Although the three-dimensional data space 36 is a three-dimensional space, it is represented in a plan view in FIG. 6 for explanation.

  As shown in FIG. 6, when a plurality of rays are set in an oblique direction intersecting with the x-axis and the y-axis, the coordinates at which each ray first enters the three-dimensional data space 36 when attention is given to each ray. Is indicated by a black circle in FIG. Each black circle corresponds to voxel data at the same time. As shown in FIG. 6, the initial coordinates that enter the three-dimensional data space 36 are different depending on the position of each ray. For example, for the eight rays from the rays L5 to L12, the approach positions exist on the yz plane that crosses the origin in the x coordinate, and are represented by voxel data e1 to e8. On the other hand, the rays L1 to L4 enter xz crossing the end in the y direction, and the voxel data at the first entry position is represented by e9 to e12 in FIG. Incidentally, one ray is associated with each coordinate on the screen 40, and in the example of FIG. 6, each ray is parallel, and for convenience, a unit ray vector V is represented on the ray L1. Yes.

  In the apparatus of the present embodiment, a plurality of voxel data (that is, volume data) existing in the three-dimensional data space 36 are transferred in units of slice data as described above, and such transfer unit is D1, It is represented by D2. That is, for each coordinate on the x axis, the plane perpendicular to the x axis is used as a transfer unit, and the data is sequentially transferred in either the positive direction or the negative direction in the x direction. In the example shown in FIG. 6, the slice data extraction direction is defined in the positive direction in the x direction. This is because the component Δx on the x-axis of the unit vector V has a positive sign as will be described later.

  In the present embodiment, when slice data D1 is transferred, voxel computation is executed for each ray using voxel data e1 to e8 only for rays L5 to L12. In this case, the voxel calculation is postponed for L1 to L4. That is, there is no voxel data that can be subjected to the voxel operation on those rays. Next, when the slice data D2 is transferred, the voxel operation is executed again for the rays L5 to L12. In addition, the voxel operation is executed for the ray L4 using the voxel data e9 for the first time. Similarly, when slice data is transferred, a ray capable of executing the voxel operation is specified among all the rays, and the voxel operation proceeds to the coordinates where the voxel operation can be executed only for that ray. Accordingly, the progress of the voxel calculation is determined independently for each ray depending on the transfer data.

  FIG. 7 shows the above items step by step, and the process proceeds in the order of (a), (b), (c), (d). That is, when the first data transfer is performed as shown in (a), the voxel operation is executed for the rays L5 to L12 using the voxel data string E1, and the voxel operations for the other rays L1 to L4 are performed. I will be sent off. Next, when data transfer is performed, as shown in (b), a voxel operation is executed for the rays L4 to L12 using the voxel data string E2. That is, a ray L4 is added as a ray capable of executing the voxel operation compared to the case shown in FIG.

  Next, when data transfer is performed as shown in (c), a voxel operation is performed on the rays L4 to L12 using the voxel data string E3, and at this stage, a ray that can execute a new voxel operation is added. Not. Then, as shown in (d), when further data transfer is performed, the voxel operation is executed for the rays L3 to L12 using the voxel data string E4. When such a process is repeated, finally all necessary voxel calculations can be performed for all the rays, and as a result, a three-dimensional image can be constructed on the screen 40. By the way, as the termination condition of the voxel calculation for each ray, for example, the target coordinates to be subjected to the voxel calculation are out of the three-dimensional data space 36, the output value exceeds the predetermined value, and the cumulative addition is not performed. It is set that the transparency value has exceeded a predetermined value, and the voxel calculation is individually ended for each ray when the end condition is satisfied.

  By the way, in the example shown in FIG. 7, the direction of each ray is determined from the upper left to the lower right direction. On the contrary, for example, the direction of each ray seems to be set from the upper right to the lower left direction. In such a case, the slice data is sequentially read in the negative direction in the x direction. Thus, the voxel calculation is appropriately advanced for each ray along the ray direction. Further, when the direction of each ray coincides with, for example, the y direction or the z direction, when one slice data is transferred, all voxel operations are completed for the ray penetrating the slice data. That is, in the present embodiment, it is a condition that the voxel operation is advanced to a place where the voxel operation can be executed for the ray in which the voxel operation can be executed in the entire ray, and thereby the image processing according to the transferred data is performed. Is realized.

  Incidentally, when the ray direction coincides with the y direction or the z direction, it is possible to specify either the positive direction or the negative direction in the x direction as the slice data extraction direction.

  FIG. 8 shows a state in which a plurality of rays intersect with the y axis at a minute angle. Even in such a case, as shown in (a), when the first data transfer is performed, the voxel calculation sequentially proceeds along the ray L1 using the voxel data string E1, and the processing for the ray L1 is performed. Complete. Then, when data is transferred next time, the voxel operation for the ray L2 is sequentially executed using the voxel data string e2, thereby completing the voxel operation for the ray L2. The same applies to the cases shown in (c) and (d). That is, volume rendering for the ray L3 is executed using the voxel data string E3, and volume rendering is executed for the ray L4 using the voxel data string E4.

  Therefore, according to the present embodiment, even when the viewpoint is set at an arbitrary position with respect to the three-dimensional data space, if slice data is read from either the positive direction or the negative direction on the x-axis, Thus, the volume rendering operation for each ray can be appropriately performed.

  When voxel calculation is performed on the target coordinates on each ray and there is no voxel data corresponding to the target coordinates, the target coordinates are interpolated using a plurality of voxel data existing in the vicinity of the target coordinates. It is desirable to generate interpolation data corresponding to the above and perform voxel calculation using the generated interpolation data. In this case, the interpolation calculation may use voxel data existing on the transferred slice data, or a process such as 4-point interpolation or 16-point interpolation using a plurality of slice data arranged on the X axis. Interpolation data may be generated using. Further, as shown in FIG. 8, the voxel data closest to the target coordinate may be used as the voxel data of the target coordinate. This is also an interpolation process in a broad sense.

  In the present embodiment, the basic length in the ray direction is defined by the unit ray vector, and the basic length is set as the interval of the target coordinates, that is, the pitch. According to such a method, the number or density of voxel operations can be made uniform regardless of the direction of each ray, so that the image quality that occurs when the pitch of voxel operations is indefinite when the viewpoints are different. It is possible to prevent problems such as differences in image quality or changes in image content. At this time, each component of the unit ray vector described above is used as described later.

  Next, the operation of the ultrasonic diagnostic apparatus shown in FIG. 2, particularly the operation of the 3D processor, will be described with reference to FIGS. FIG. 9 shows the main routine. In S101, the initialization process shown in FIG. 10 is executed. In S102, the slice data extraction direction is selected using the sign of the component ray vector component Δx, that is, whether to execute the process of S103 or the process of S104 is selected.

  In S103, the positive direction in the x direction is designated as the extraction direction of slice data, and image processing according to the data transfer is executed. This will be described in detail later with reference to FIG. On the other hand, in S104, the negative direction in the x direction is designated as the data extraction method, and image processing based on that is executed. The process of S103 and the process of S104 are basically the same, but some parameters are set there. This will be described later. In S105, the 3D image generated by any of the above processes is transferred from the 3D processor to the display processing unit.

  FIG. 10 is a flowchart showing the initialization process of S101 shown in FIG.

  First, in S201, components for each axis, that is, elements Δx, Δy, and Δz are calculated for the normalized line-of-sight (unit ray) vector. The unit ray vector has a length of 1 and its direction is defined as the direction of viewing the screen from the viewpoint. In S202, 0 is substituted for the parameter Y, and 0 is substituted for the parameter X in S203.

In S204, for the ray specified by the screen coordinates X and Y, the point (x _s , y _s , z _s ) at which it first intersects with the three-dimensional data space is calculated as the target coordinate to be subjected to the first voxel calculation. The In step S205, the target coordinates are stored in an address table configured on the internal memory.

  In S206, the parameter X is incremented by one. In S207, it is determined whether or not X exceeds the maximum value. If not, each process from S204 is executed.

  On the other hand, in S208, the parameter Y is incremented by 1, and in S209, it is determined whether or not the parameter Y has exceeded its maximum value. If not, each process from S203 is repeatedly executed. That is, the first target coordinates are calculated for each ray while changing the parameters X and Y one by one, and stored in the address table.

  FIG. 11 shows a specific content of the process of S103 shown in FIG. 9 as a flowchart.

In S301, 0 is substituted for the parameter x (in the case of S104 shown in FIG. 9, x = _xmax is substituted). In S302, slice data is transferred. That is, voxel data specified by x, y: 0 to y _max and z: 0 to z _max is transferred from the external memory to the internal memory as a block. The data set constitutes the above-described slice data, that is, a data set continuous in an actual memory space. By using such a data set as a transfer unit, the conventional random access becomes unnecessary, and there is an advantage that necessary data can be transferred collectively. Incidentally, 0 is substituted for the parameter x in S301, and x is incremented by 1 as will be described later, so that the slice data is sequentially extracted in the positive direction in the x direction.

  In S303, 0 is substituted for the parameter Y, and in S304, 0 is substituted for the parameter X. In S305, it is determined whether or not the ray specified by X and Y is a ray for which all voxel operations have been completed. If Yes, each step from S312 is executed, and if No, each step from S306 is executed.

In S306, the target coordinates (x _p , y _p , z _p ) are read from the address table for the ray specified by X and Y. That is, the coordinates currently being calculated for the ray are specified. Then, in S307, x _p is whether they match the x is determined, the process as voxel data does not exist that can be operational in must match currently transferred come within the slice data is shifted to S312 On the other hand, if they match, the process proceeds to S308 on the assumption that the voxel data to be subjected to the voxel calculation is present in the slice data.

  In S308, the output value is calculated by executing the above equation (1) using the voxel data specified by the target coordinates.

In S309, the calculation result is stored in a table on the internal memory. That is, according to the above equation (1), C _outXY and α _outXY which are the calculation results for the _ray are stored. Incidentally, in this stage, when a predetermined end condition is satisfied, for example, when the accumulated addition value of the opacity value reaches 1, for example, the process proceeds to S312 before the data is stored or after the data is stored without moving the process to S310. It is desirable to migrate.

  In S310, the target coordinates are updated for the ray. That is, Δx, Δy, and Δz are added for each of the x coordinate, the y coordinate, and the z coordinate, thereby updating the target coordinate.

  In S311, it is determined whether or not such updated target coordinates are in the three-dimensional data space (or effective processing range). If the target coordinates are in the three-dimensional data space, the process proceeds to S307. On the other hand, if the target coordinates are not within the three-dimensional data space, the process proceeds to S312.

  Here, when the process proceeds from S311 to S307, if there is further voxel data that can be subjected to voxel calculation in the slice data currently being processed, the processes of S308 are repeatedly performed. Become. That is, by such a process. For each slice data, that is, for each data transfer, a ray capable of executing the voxel operation is determined, and the voxel operation can be advanced to a place where the voxel operation can be performed for the ray.

  In S312, the parameter X is incremented by one. In S313, it is determined whether or not the parameter X exceeds the maximum value. If not, the process proceeds to S305. If it exceeds, the parameter Y is set in S314. It is incremented by 1, and it is determined in S315 whether or not the parameter Y exceeds the maximum value. If Y does not exceed the maximum value, the process proceeds to S304. If Y exceeds the maximum value, the process proceeds to S316. In S316, x is incremented by 1 (x = x-1 in the case of S104 in FIG. 9), and in S317, it is determined whether or not the parameter x exceeds the maximum value (in the case of S104 in FIG. 9). X <0 is determined), if not, the process proceeds to S302, and if it exceeds, the process proceeds to S318.

That is, for each transfer of slice data, it is determined whether or not the execution of the voxel operation is possible for each ray, and if it is possible, the voxel operation proceeds to the point where the voxel operation can be executed. Then, after such verification and calculation are performed for all the rays, the process returns to S302 to transfer the next slice data, and this is sequentially repeated for each transfer of slice data. Therefore, when the transfer from the first slice data to the last slice data is completed, all voxel operations that can be executed for all the rays are completed as a result. Therefore, in S108, a process of performing luminance conversion on C _outXY , which is an output value obtained for each ray, and storing it as a pixel value is executed.

  Next, a modification of the above embodiment will be described with reference to FIGS.

  Although one 3D memory 24 and one 3D processor 26 are shown in FIG. 2, a plurality of each may be provided. That is, as shown in FIG. 12, a plurality of 3D processors 26A, 26B, and 26C may be provided for the 3D memory 24. Each processor 26A, 26B, 26C is in charge of each group grouped for a plurality of rays. The slice data is transferred in parallel to each processor 26A, 26B, 26C.

  Similarly, a configuration as shown in FIG. 13 can be adopted. That is, three 3D memories 24A, 24B, and 24C are provided, and the same volume data is redundantly stored for them. Three 3D processors 26A, 26B, and 26C are provided, which advance voxel operations for the rays in the group that they are responsible for according to the slice data output from the 3D memories 24A, 24B, and 24C.

  According to the configuration example shown in FIG. 12 and FIG. 13, it is possible to further improve the image processing speed by operating a plurality of 3D processors in parallel and sharing a plurality of rays to advance a voxel operation. In this case, it is desirable that the data transfer from the 3D memory to the 3D processor is exactly the same for each 3D processor.

  14 and 15 show a grouping method for a plurality of rays, and FIG. 14A shows a method in which a plurality of rays are allocated to three groups in a predetermined sequence with one ray as a unit. ing. For example, pixels 50, 51, and 52 are associated with three rays arranged side by side. According to such a method, the pixel value can be calculated for each ray in a pattern as shown in FIG. Further, as shown in FIG. 14B, grouping may be performed in units of lines on the screen 40. That is, a 3D processor in charge of each of the lines 53, 54, and 55 is assigned. Grouping is performed in a pattern as shown in FIG.

  In the example shown in FIG. 14C, the screen 40 is divided into a plurality of areas 56, 57, and 58, and a 3D processor in charge is assigned to each area. As a result, grouping is performed with a pattern as shown in FIG.

  As described above, according to the above-described embodiment, transfer to the 3D processor can be performed in units of data sets in which memory addresses are continuous on the 3D memory, so that transfer control is simple and transfer is possible. Time can be significantly reduced. In that case, it is desirable that the burst transfer method is adopted and the DMA transfer is performed by the DMA controller existing in the 3D processor. Here, in general, the width of the bus between the 3D processor and the external memory is very wide, for example, 32 bits. When the voxel data is 8 bits, for example, if the voxel data is read by random access, the above wide bus cannot be used effectively, and the data can be read only 8 bits at a time. On the other hand, according to the above embodiment, it is possible to read a plurality of voxel data in parallel. For example, if the bus width is 32 bits, parallel reading of four voxel data can be realized, and the bus width is 64 bits. If there is, there is an advantage that eight voxel data can be read in parallel. From this point of view, there is an advantage that the data transfer time can be shortened.

It is explanatory drawing for demonstrating the principle of volume rendering. 1 is a block diagram showing an overall configuration of an ultrasonic diagnostic apparatus according to the present embodiment. It is a figure which shows notionally the structure of three-dimensional data space. It is a figure which shows the address structure in 3D memory. It is explanatory drawing for demonstrating a unit ray vector. It is a figure which shows the relationship between slice data and the voxel calculation for every ray. It is explanatory drawing which showed a mode that each slice data was utilized by voxel calculation in steps. It is explanatory drawing which showed a mode that each slice data was utilized by voxel calculation in steps. It is a flowchart which shows the operation example of an apparatus. It is a flowchart which shows the specific content of the initialization process shown in FIG. It is a flowchart which shows the specific content of the process of S103 shown in FIG. It is a figure which shows the structural example in the case of providing one 3D memory and three 3D processors. It is a figure which shows the structural example in the case of providing three 3D memories and three 3D processors. It is explanatory drawing for demonstrating grouping. It is explanatory drawing for demonstrating grouping.

Explanation of symbols

  10 3D probe, 12 transmit / receive space, 22 3D converter, 24 3D memory (external memory), 26 3D processor, 28 internal memory, 36 3D data space, 40 screens.

Claims (12)

A transmission / reception means for transmitting / receiving ultrasonic waves to / from a living body, thereby capturing volume data composed of a plurality of voxel data;
A three-dimensional data memory in which the volume data is stored;
Three-dimensional image forming means for forming a three-dimensional image based on the volume data;
Means for transferring the volume data from the three-dimensional data memory to the three-dimensional image forming means, wherein the data transfer means repeatedly executes data transfer with a predetermined data set as a transfer unit;
Including
The three-dimensional image forming means includes:
Ray setting means for setting a plurality of rays for the three-dimensional data space;
Ray determination means for determining a ray capable of executing voxel calculation among the plurality of rays for each data transfer;
Voxel computing means for advancing the voxel computation to the coordinates at which the voxel computation can be performed for each data transfer;
Including
The 3D image is constructed as a result of the last voxel operation for each ray,
An ultrasonic diagnostic apparatus, wherein voxel calculation proceeds independently for each ray. The apparatus of claim 1.
The predetermined data set is slice data composed of a plurality of voxel data having continuity in a three-dimensional data space,
The ultrasonic diagnostic apparatus, wherein the data transfer means repeatedly executes data transfer in the order of slice data. The apparatus of claim 2.
The ultrasonic diagnostic apparatus, wherein the data transfer unit includes a selection unit that selects a positive direction or a negative direction in the arrangement direction of the slice data as the extraction direction of the slice data. The apparatus of claim 3.
The ultrasonic diagnostic apparatus, wherein the direction selection unit selects the forward direction or the reverse direction according to a position of a viewpoint with respect to the three-dimensional data space. The apparatus of claim 2.
Coordinate conversion means for executing coordinate conversion from the coordinate system of the three-dimensional transmission / reception space to the coordinate system of the three-dimensional data space when the volume data is written to the three-dimensional data memory;
The data storage space has three orthogonal coordinate axes,
The ultrasonic diagnostic apparatus according to claim 1, wherein the slice data corresponds to a plane orthogonal to a specific coordinate axis among the three orthogonal coordinate axes. The apparatus of claim 1.
The ultrasonic diagnostic apparatus, wherein the three-dimensional image forming means includes data interpolation means for generating data necessary for the voxel calculation by interpolation processing. The apparatus of claim 1.
The ultrasonic diagnostic apparatus, wherein the three-dimensional image forming means includes means for equalizing the number of voxel operations per unit length for each ray. The apparatus of claim 1.
The three-dimensional image forming means includes a plurality of three-dimensional image forming modules operating in parallel,
The plurality of rays are divided into a plurality of groups;
The ultrasonic diagnostic apparatus, wherein each of the three-dimensional image forming modules is in charge of voxel calculation for a group associated therewith. A transmission / reception means for transmitting / receiving ultrasonic waves to / from a three-dimensional transmission / reception space of a living body, thereby capturing volume data composed of a plurality of voxel data;
A three-dimensional data space having three coordinate axes orthogonal to each other and storing the volume data;
Three-dimensional image forming means for forming a three-dimensional image based on the volume data;
Means for transferring the volume data from the three-dimensional data memory to the three-dimensional image forming means, wherein slice data orthogonal to specific coordinate axes in the three coordinate axes in the three-dimensional data space is a transfer unit; and Data transfer means for repeatedly executing data transfer, with the direction of the specific coordinate axis as the slice data extraction direction;
Including
The three-dimensional image forming means includes:
Ray setting means for setting a plurality of rays for the three-dimensional data space;
An internal memory for temporarily storing the sliced data transferred;
Each time the data is transferred, the slice data stored in the internal memory is used to advance the voxel operation to the coordinates at which the voxel operation can be executed among the plurality of rays up to the coordinates at which the voxel operation can be executed. Means for determining a pixel value constituting the three-dimensional image as a result of the last voxel operation for each ray;
An ultrasonic diagnostic apparatus comprising: The apparatus of claim 9.
The three-dimensional image forming means further includes:
For each ray, as the first point it passes through the 3D data space, the initial coordinates on the first axis, the initial coordinates on the second axis and the 3rd axis on the 3D data space. Initial coordinate calculating means for calculating initial coordinates;
A step component calculating means for calculating a step component of the first axis, a step component of the second axis, and a step component of the third axis;
For each ray, for each initial coordinate on the first axis, initial coordinate on the second axis, and initial coordinate on the third axis, the step component on the first axis and the second axis, respectively. Target coordinate specifying means for specifying the voxel calculation target coordinates on each ray by cumulatively adding the step component on the upper axis and the step component on the third axis;
The voxel data corresponding to the voxel calculation target coordinates is extracted from the slice data stored in the internal memory, or the slice data stored in the internal memory is used to correspond to the voxel calculation target coordinates. Means for generating voxel data by interpolation;
An ultrasonic diagnostic apparatus comprising: The apparatus of claim 9.
Means for obtaining a coordinate component on the specific coordinate axis for a unit ray vector defined for the three-dimensional data space;
When the coordinate component on the specific coordinate axis is positive, the positive direction of the specific coordinate axis is selected as the slice data extraction direction, and when the coordinate component of the specific coordinate axis is negative, the negative direction of the specific coordinate axis is selected. Means for selecting the direction as the extraction direction of slice data;
An ultrasonic diagnostic apparatus comprising: A transfer step of transferring the volume data from a three-dimensional data memory storing volume data composed of a plurality of voxel data to an internal memory in the image processor;
An image forming step of forming a three-dimensional image based on a volume rendering method using the transferred volume data;
In the volume data processing method including
The transfer step includes a step of dividing the volume data into a plurality of slice data aligned in a specific coordinate axis direction, and sequentially transferring the slice data from the plurality of slice data in the direction of the predetermined coordinate axis,
The image forming step includes
Setting a plurality of rays for the three-dimensional data space;
For each of the data transfers, the voxel operation is advanced to the coordinates at which the voxel operation can be performed among the rays in which the voxel operation can be performed among the plurality of rays, and as a result of the last voxel operation for each ray, the three-dimensional Obtaining pixel values constituting an image;
A volume data processing method comprising:

Images