CN107577642A - A kind of shear-wave direction filter achieving method and medical supersonic wave device - Google Patents
A kind of shear-wave direction filter achieving method and medical supersonic wave device Download PDFInfo
- Publication number
- CN107577642A CN107577642A CN201710744330.5A CN201710744330A CN107577642A CN 107577642 A CN107577642 A CN 107577642A CN 201710744330 A CN201710744330 A CN 201710744330A CN 107577642 A CN107577642 A CN 107577642A
- Authority
- CN
- China
- Prior art keywords
- operator
- filtering
- matrix
- data
- wave
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Processing (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The invention discloses a kind of shear-wave direction filter achieving method and medical supersonic wave device, this method includes structural grain gating operator, two-dimension fourier inverse transformation, construction convolutional filtering operator is carried out to direction gating operator, carries out the steps such as parallel-convolution filtering operation to every frame data.The ultrasonic equipment includes the device of shear-wave direction filter achieving method.By the way that convolutional filtering method is replaced into traditional frequency domain spatially selecting filtering method in the present invention, overcome two-dimensional Fourier transform and be unable to the problem of parallel processing causes computing time-consuming and committed memory space is double.Convolutional filtering method carries out convolution algorithm by using filter operator to pending data, reaches the effect of trend pass filtering.This process parallel processing and can will not increase EMS memory occupation space.
Description
Technical field
The present invention relates to shearing wave elastogram field, more particularly to a kind of shear-wave direction filter achieving method and use
Direction filter achieving method carries out the medical supersonic wave device of image procossing.
Background technology
Shearing wave is the direction of propagation ripple vertical with the direction of vibration of medium particle, also known as shear wave, S ripples.
Medical supersonic wave device can produce ultrasonic excitation tissue shearing wave by supersonic generator and propagate, by
In different tissue densities, shearing velocity of wave propagation has difference, so may determine that shape of different piece etc. in tissue, this is just
It has been the principle of medical shearing wave imaging.The effect of trend pass filtering is a certain specific in order to obtain in medical science shearing wave elastogram
The shearing wave information that direction is propagated.And the implementation method of general direction filtering is by carrying out two to shearing wave Displacement Estimation matrix
Dimension Fourier transformation method draws frequency domain information, and after being multiplied with direction gating operator, filtering a direction frequency domain information, then enter
Row two-dimensional inverse Fourier transform obtains trend pass filtering result.The shortcomings that the method is when carrying out two-dimensional Fourier transform, and row becomes
It is a serial step to change with rank transformation, and similarly two-dimensional inverse Fourier transform is also such.But shearing wave two dimension is elastic in real time
In imaging, signal transacting data volume is huge, requires very high to efficiency of algorithm, and this method calculates to take to be grown very much, even if knot
Close the parallel acceleration techniques of GPU, due to the inside two-dimensional Fourier transform can not parallelization and cause trend pass filtering accelerate to imitate parallel
Rate is very low.
The content of the invention
The shortcomings that for above-mentioned trend pass filtering implementation method, the present invention propose a kind of shear-wave direction filter achieving method with
And the medical supersonic wave device using this method, this method replaces Fourier transformation by using convolutional calculation method, and combines
GPU parallel optimization techniques, greatly reduce trend pass filtering and take, and result of calculation is consistent with Fourier transformation method.
The present invention is that technical scheme is used by realizing its purpose:A kind of shear-wave direction filter achieving method, including
Following steps:.
Step 1, structural grain gating operator;In the step, according to pending data size and required filtering directional structure vectorical structure
Frequency domain direction gates operator;
Step 2, two-dimension fourier inverse transformation is carried out to frequency domain direction gating operator;
Step 3, construction convolutional filtering operator;In the step, the real part data of the result of selecting step 2 are filtered as convolution
Wave operator;
It is step 4, parallel to being carried out in pending shearing wave Displacement Estimation matrix per frame data using convolutional filtering operator
Convolutional filtering computing.
The present invention proposes the trend pass filtering implementation method that a kind of combination GPU optimization is realized, this method is by by convolutional filtering
Method replaces traditional frequency domain spatially selecting filtering method, and overcoming two-dimensional Fourier transform and being unable to parallel processing causes computing time-consuming
With committed memory space it is double the problem of.Convolutional filtering method carries out convolution fortune by using filter operator to pending data
Calculate, reach the effect of trend pass filtering.This process parallel processing and can will not increase EMS memory occupation space.
The convolutional filtering method of the present invention is not due to having series steps during convolution algorithm, and convolution algorithm is interframe
Independent, you can maximal efficiency optimizes calculating using GPU.
Further, in above-mentioned shear-wave direction filter achieving method:It is Row*Line* for pending data size
During Frame, Row represents number of data lines herein, and Line represents data columns, and Frame represents data frame number, and trend pass filtering computing should
Data matrix for the Row*Line sizes of each frame;The matrix that described direction gating operator size is Row*Line,
Building method is as follows:
Dir_Mask_L represents to choose from left to right direction in above formula, and Dir_Mask_R represents direction from right to left.
Further, in above-mentioned shear-wave direction filter achieving method:In described step 3, comprise the following steps:
The matrix data after two-dimension fourier inverse transformation is carried out to step 2 direction gating operator and carries out above-below direction upset;
Left and right directions upset is carried out to matrix of consequence after above-mentioned upset again;
The real part of matrix of consequence data is finally taken as convolutional filtering operator matrix.
The present invention also provides a kind of medical supersonic wave device, including:
Produce the ultrasonic generator that Ultrasonic Radiation power shock-excitation shearing wave passes through biological tissue's viewing area;
The Hyperframe of biological tissue's viewing area shearing wave Displacement Estimation matrix is gathered using superelevation frame frequency imaging technique
Frequency imaging device;
Biological tissue's viewing area shearing wave Displacement Estimation matrix of superelevation frame frequency imaging device output is sheared
The device of ripple trend pass filtering;It is characterized in that:Described biological tissue's viewing area to the output of superelevation frame frequency imaging device
The device that shearing wave Displacement Estimation matrix carries out shear-wave direction filtering includes:
According to shearing wave Displacement Estimation matrix size and the frequency domain of required filtering directional structure vectorical structure frequency domain direction gating operator
Direction gates operator constructing apparatus,
Two-dimension fourier inverse transformation is carried out to the frequency domain direction gating operator of frequency domain direction gating operator constructing apparatus construction
Form the two-dimension fourier inverse transformation device of time domain direction gating operator;
Dress is constructed from convolutional filtering operator of the time domain direction gating operator real part data as convolutional filtering operator
Put;
Using convolutional filtering operator to carrying out parallel-convolution filter in pending shearing wave Displacement Estimation matrix per frame data
The device of ripple computing.
Further, in above-mentioned medical supersonic wave device:Described frequency domain direction gating operator constructing apparatus, construction is such as
Lower frequency domain gates operator:
Dir_Mask_L represents to choose from left to right direction in above formula, and Dir_Mask_R represents direction from right to left.
Further, in above-mentioned medical supersonic wave device:Described convolutional filtering operator constructing apparatus includes:
The first turning device spun upside down to time domain direction gating operator;
The second turning device of left and right upset is carried out to matrix caused by the first described turning device;
The real part for the matrix that the second described turning device exports is taken to form the 3rd device of convolutional filtering operator matrix.
The present invention is described in more detail with reference to specific embodiment.
Brief description of the drawings
Fig. 1 is the flow chart of the embodiment of the present invention 1.
Fig. 2 is frequency domain gating matrix schematic diagram.
Fig. 3 spins upside down schematic diagram.
Fig. 4 or so overturns schematic diagram.
Fig. 5 is convolutional filtering operator figure.
Before Fig. 6 is trend pass filtering.
After Fig. 7 is trend pass filtering.
Embodiment
Embodiment 1, the present embodiment are a kind of medical supersonic wave devices, are had in the device:
Produce the ultrasonic generator that Ultrasonic Radiation power shock-excitation shearing wave passes through biological tissue's viewing area.
The Hyperframe of biological tissue's viewing area shearing wave Displacement Estimation matrix is gathered using superelevation frame frequency imaging technique
Frequency imaging device.
Biological tissue's viewing area shearing wave Displacement Estimation matrix of superelevation frame frequency imaging device output is sheared
The device of ripple trend pass filtering.
Output data size is Row row Line row Frame frames (Row*Line*Frame) in superelevation frame frequency imaging device,
Data type is float32.
Biological tissue's viewing area shearing wave Displacement Estimation matrix of superelevation frame frequency imaging device output is sheared
The device of ripple trend pass filtering filters to above-mentioned data travel direction.
The tool of shear-wave direction filtering stops step with reference to shown in figure 1, comprises the following steps:
Step 1, reference method flow chart, step 1 are according to pending data size and required filtering directional structure vectorical structure direction
Gate operator.If pending data size is Row*Line*Frame, Row represents number of data lines herein, and Line represents data row
Number, Frame represent data frame number.Trend pass filtering computing is applied on the data matrix of the Row*Line sizes of each frame, therefore
The matrix that direction gating operator size is Row*Line, building method are as follows:
In above-mentioned direction gating operator matrix, Dir_Mask_L is represented to choose from left to right direction, and Dir_Mask_R is represented
Direction from right to left.Wherein Dir_Mask_L operators internal structure is:By the operator matrix quartering, the 1st and 3 quadrant range sections
Value be 1, remainder 0.As shown in Figure 2, black portions represent 0 value to its result, and white portion is expressed as 1.Dir_
Mask_L is a Central Symmetry operator.Dir_Mask_R operator building methods are the same, and data value is opposite.Its result is as shown in Figure 2
It is rotated by 90 °.
In the present embodiment, according to actual requirement, the value among 0~1 is selected in the value that construction is gated in operator matrix in operator,
Represent trend pass filtering selection weighted value.Weighted value is bigger to represent that selection direction information is more, otherwise chooses direction information and get over
It is few.
Step 2, two-dimensional inverse Fourier transform is carried out, but be with the difference of conventional direction filtering method, two dimension herein
Fourier inversion is applied in the gating operator of direction, rather than in data matrix.Operator is gated to the direction constructed in step 1
Matrix carries out two-dimensional inverse Fourier transform.Two-dimensional inverse Fourier transform method is consistent with conventional method, first gates operator to direction
Matrix along column direction to each column data carry out Fourier inversion, then to matrix of consequence along line direction to every data line
Carry out Fourier inversion.Because the direction used in all frames gates operator, whole process only needs to carry out once
Two-dimensional Fourier transform computing.
Step 3, it is that convolutional filtering operator constructs, constitution step is as follows:
Above-below direction upset is carried out to step 2 matrix of consequence data first, as shown in Figure 3.
Then left and right directions upset is carried out to matrix of consequence data after upset again, as shown in Figure 4.
The real part of matrix of consequence data is finally taken as final convolutional filtering operator matrix.
Step 4, the convolutional filtering operator with step 3, two-dimensional convolution filtering operation is carried out to each frame pending data.By
It is interframe independence in convolution process, and the processing method of each frame and the convolution operator that uses are all.Here with GPU simultaneously
Row treatment technology, to the data matrix that size is Row*Line*Frame, divided in units of frame Frame so that a thread
The convolution algorithm of a frame is completed, the convolution algorithm of all frames is synchronously carried out.Utilize the parallel acceleration techniques of GPU so that N frame data
Processing time and 1 frame approach, and speed-up ratio reaches N times.
In the present embodiment, because volume computing substitutes inverse fourier transform, many times and memory space can be saved.Assuming that
Pending data size is Row row Line row Frame frames (Row*Line*Frame), data type float32.Contrast respectively
The time-consuming and EMS memory occupation situation of conventional direction filtering algorithm and trend pass filtering algorithm of the present invention, two methods use GPU skills
Art, division progress parallel processing acceleration, test data size select Row=500, Line=100, Frame=in units of frame
100:Specifically it is shown in Table 1
Table 1
For the method for the present embodiment compared with conventional method, speed-up ratio reaches 12 times, and memory headroom needed for conventional method
For twice of pending data size, and memory headroom needed for the inventive method and pending data are in the same size.So this reality
The method either efficiency of algorithm or space complexity for applying example will be more excellent than conventional direction filtering algorithm, are more suitable for this reality
Apply the medical supersonic wave device in example.
Claims (6)
- A kind of 1. shear-wave direction filter achieving method, it is characterised in that:Comprise the following steps:.Step 1, structural grain gating operator;In the step, according to pending data size and required filtering directional structure vectorical structure frequency domain Direction gates operator;Step 2, two-dimension fourier inverse transformation is carried out to frequency domain direction gating operator;Step 3, construction convolutional filtering operator;In the step, the real part data of the result of selecting step 2 are calculated as convolutional filtering Son;Step 4, using convolutional filtering operator in pending shearing wave Displacement Estimation matrix per frame data carry out parallel-convolution Filtering operation.
- 2. shear-wave direction filter achieving method according to claim 1, it is characterised in that:For pending data size For Row*Line*Frame when, herein Row represent number of data lines, Line represent data columns, Frame represent data frame number, side It is applied to filtering operation on the data matrix of the Row*Line sizes of each frame;Described direction gating operator size is Row* Line matrix, building method are as follows:Dir_Mask_L represents to choose from left to right direction in above formula, and Dir_Mask_R represents direction from right to left.
- 3. shear-wave direction filter achieving method according to claim 1, it is characterised in that:In described step 3, including Following steps:The matrix data after two-dimension fourier inverse transformation is carried out to step 2 direction gating operator and carries out above-below direction upset;Left and right directions upset is carried out to matrix of consequence after above-mentioned upset again;The real part of matrix of consequence data is finally taken as convolutional filtering operator matrix.
- 4. a kind of medical supersonic wave device, including:Produce the ultrasonic generator that Ultrasonic Radiation power shock-excitation shearing wave passes through biological tissue's viewing area;Using superelevation frame frequency imaging technique gather the superelevation frame frequency of biological tissue's viewing area shearing wave Displacement Estimation matrix into As device;Shearing wave side is carried out to biological tissue's viewing area shearing wave Displacement Estimation matrix of superelevation frame frequency imaging device output To the device of filtering;It is characterized in that:Described biological tissue's viewing area shearing to the output of superelevation frame frequency imaging device The device that ripple Displacement Estimation matrix carries out shear-wave direction filtering includes:According to shearing wave Displacement Estimation matrix size and the frequency domain direction of required filtering directional structure vectorical structure frequency domain direction gating operator Operator constructing apparatus is gated,The frequency domain direction gating operator progress two-dimension fourier inverse transformation of frequency domain direction gating operator constructing apparatus construction is formed Time domain direction gates the two-dimension fourier inverse transformation device of operator;From convolutional filtering operator constructing apparatus of the time domain direction gating operator real part data as convolutional filtering operator;Using convolutional filtering operator to carrying out parallel-convolution filtering fortune in pending shearing wave Displacement Estimation matrix per frame data The device of calculation.
- 5. medical supersonic wave device according to claim 4, it is characterised in that:Described frequency domain direction gating operator construction Device, construct following frequency domain gating operator:Dir_Mask_L represents to choose from left to right direction in above formula, and Dir_Mask_R represents direction from right to left.
- 6. medical supersonic wave device according to claim 4, it is characterised in that:Described convolutional filtering operator constructing apparatus Including:The first turning device spun upside down to time domain direction gating operator;The second turning device of left and right upset is carried out to matrix caused by the first described turning device;The real part for the matrix that the second described turning device exports is taken to form the 3rd device of convolutional filtering operator matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710744330.5A CN107577642B (en) | 2017-08-25 | 2017-08-25 | Shear wave direction filtering implementation method and medical ultrasonic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710744330.5A CN107577642B (en) | 2017-08-25 | 2017-08-25 | Shear wave direction filtering implementation method and medical ultrasonic equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107577642A true CN107577642A (en) | 2018-01-12 |
CN107577642B CN107577642B (en) | 2020-06-23 |
Family
ID=61035419
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710744330.5A Active CN107577642B (en) | 2017-08-25 | 2017-08-25 | Shear wave direction filtering implementation method and medical ultrasonic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107577642B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109512465A (en) * | 2018-12-19 | 2019-03-26 | 深圳中科乐普医疗技术有限公司 | A kind of acoustic radiation force double direction shear wave composite imaging method and device |
CN110742644A (en) * | 2019-09-29 | 2020-02-04 | 深圳大学 | Elastography system, elastography method and storage medium |
WO2023071528A1 (en) * | 2021-10-27 | 2023-05-04 | 青岛海信医疗设备股份有限公司 | Shear wave propagation speed determination method and ultrasonic device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140368494A1 (en) * | 2013-06-18 | 2014-12-18 | Nvidia Corporation | Method and system for rendering simulated depth-of-field visual effect |
CN106618635A (en) * | 2017-01-12 | 2017-05-10 | 清华大学 | Shear wave elastic imaging method and device |
CN106934110A (en) * | 2016-12-14 | 2017-07-07 | 北京信息科技大学 | A kind of filtered back-projection method and device that light field is rebuild by focusing storehouse |
-
2017
- 2017-08-25 CN CN201710744330.5A patent/CN107577642B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140368494A1 (en) * | 2013-06-18 | 2014-12-18 | Nvidia Corporation | Method and system for rendering simulated depth-of-field visual effect |
CN106934110A (en) * | 2016-12-14 | 2017-07-07 | 北京信息科技大学 | A kind of filtered back-projection method and device that light field is rebuild by focusing storehouse |
CN106618635A (en) * | 2017-01-12 | 2017-05-10 | 清华大学 | Shear wave elastic imaging method and device |
Non-Patent Citations (1)
Title |
---|
G. PRASAD ACHARYA等: ""Implementation of High Speed Convolution Algorithm on CUDA based Graphics Processing Unit"", 《INTERNATIONAL JOURNAL OF COMPUTER APPLICATIONS(0957-8887)》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109512465A (en) * | 2018-12-19 | 2019-03-26 | 深圳中科乐普医疗技术有限公司 | A kind of acoustic radiation force double direction shear wave composite imaging method and device |
CN109512465B (en) * | 2018-12-19 | 2021-04-06 | 深圳中科乐普医疗技术有限公司 | Acoustic radiation force bidirectional shear wave composite imaging method and device |
CN110742644A (en) * | 2019-09-29 | 2020-02-04 | 深圳大学 | Elastography system, elastography method and storage medium |
CN110742644B (en) * | 2019-09-29 | 2022-07-08 | 深圳大学 | Elastography system, elastography method and storage medium |
WO2023071528A1 (en) * | 2021-10-27 | 2023-05-04 | 青岛海信医疗设备股份有限公司 | Shear wave propagation speed determination method and ultrasonic device |
Also Published As
Publication number | Publication date |
---|---|
CN107577642B (en) | 2020-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107577642A (en) | A kind of shear-wave direction filter achieving method and medical supersonic wave device | |
CN107729994B (en) | Method and apparatus for performing operations on convolutional layers in convolutional neural networks | |
US11645529B2 (en) | Sparsifying neural network models | |
CN107341544A (en) | A kind of reconfigurable accelerator and its implementation based on divisible array | |
CN108268931B (en) | Data processing method, device and system | |
CN108198625B (en) | Deep learning method and device for analyzing high-dimensional medical data | |
CN104881640B (en) | A kind of method and device for obtaining vector | |
CN109472356A (en) | A kind of accelerator and method of restructural neural network algorithm | |
US20190279092A1 (en) | Convolutional Neural Network Compression | |
CN107392852A (en) | Super resolution ratio reconstruction method, device, equipment and the storage medium of depth image | |
CN113689517B (en) | Image texture synthesis method and system for multi-scale channel attention network | |
CN110210524A (en) | A kind of training method, image enchancing method and the device of image enhancement model | |
JP2019102082A (en) | Method and device for performing convolution calculation to tucked feature data | |
CN109598732A (en) | A kind of medical image cutting method based on three-dimensional space weighting | |
CN111738276A (en) | Image processing method, device and equipment based on multi-core convolutional neural network | |
CN103985083B (en) | Reconfigurable one-dimensional convolution processor | |
CN108095756A (en) | A kind of super-resolution plane wave ultrasonic imaging method based on SOFI | |
KR20190048279A (en) | Image processing method and apparatus using convolution neural network | |
Esmaeilzehi et al. | SRNSSI: a deep light-weight network for single image super resolution using spatial and spectral information | |
CN106296649B (en) | A kind of texture image segmenting method based on Level Set Models | |
CN103578100A (en) | Method and apparatus for simulating depth of field (dof) in microscopy | |
CN105741289A (en) | Serialized automatic segmentation method for super-large-scale virtual human slice images | |
CN110335342B (en) | Real-time hand model generation method for immersive simulator | |
CN115393191A (en) | Method, device and equipment for reconstructing super-resolution of lightweight remote sensing image | |
Ma et al. | Back-projection-based progressive growing generative adversarial network for single image super-resolution |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 405, building 9, baiwangxin Industrial Park, No. 1002, Songbai Road, Yangguang community, Xili street, Nanshan District, Shenzhen City, Guangdong Province Applicant after: Shenzhen Zhongke Medical Technology Co., Ltd. Address before: Baoan District Shiyan street Shenzhen city Guangdong province 518000 White Pine Road in Yuntai technology industrial factory building 9 building seven floor East Applicant before: Shenzhen Zhongke Medical Technology Co., Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |