CN114548191A - Photoacoustic imaging annular sparse array signal prediction method and device - Google Patents
Photoacoustic imaging annular sparse array signal prediction method and device Download PDFInfo
- Publication number
- CN114548191A CN114548191A CN202210450980.XA CN202210450980A CN114548191A CN 114548191 A CN114548191 A CN 114548191A CN 202210450980 A CN202210450980 A CN 202210450980A CN 114548191 A CN114548191 A CN 114548191A
- Authority
- CN
- China
- Prior art keywords
- array
- signal
- ultrasonic transducer
- signals
- sparse array
- 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
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000003384 imaging method Methods 0.000 title claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000012360 testing method Methods 0.000 claims description 30
- 238000013528 artificial neural network Methods 0.000 claims description 28
- 125000004122 cyclic group Chemical group 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 11
- 238000003860 storage Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000002604 ultrasonography Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000004088 simulation Methods 0.000 abstract 1
- 238000012545 processing Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 239000013598 vector Substances 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000026683 transduction Effects 0.000 description 1
- 238000010361 transduction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/5269—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving detection or reduction of artifacts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/22—Source localisation; Inverse modelling
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- General Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Biophysics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Artificial Intelligence (AREA)
- Radiology & Medical Imaging (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The invention discloses a method and a device for predicting a photoacoustic imaging annular sparse array signal. After the sparse array data is supplemented, the reconstructed photoacoustic image can effectively inhibit artifacts caused by sparse data reconstruction. The sparse array can improve the data acquisition efficiency and save the cost of an experimental system. The photoacoustic imaging is reconstructed by the sparse array signal prediction method provided by the invention, simulation and experimental image data do not need to be collected for training, and time sequence photoacoustic signal data is adopted for training and predicting, so that the method is a simple and effective mode for reconstructing high-quality images from sparse array data.
Description
Technical Field
The invention belongs to the field of artificial intelligence and biomedical engineering, and particularly relates to a photoacoustic imaging annular sparse array signal prediction method and device.
Background
Photoacoustic imaging is an emerging biomedical imaging mode formed on the basis of respective advantages of fusion optical and acoustic technologies, and generally adopts a laser emission light source and a multi-array-element ultrasonic transducer to receive acoustic signals. The photoacoustic imaging device requires a precise manufacturing process and high-density array element devices, which greatly increases the cost and complexity of the imaging system. High density array ultrasound transducers are subject to further development subject to processing limitations. Meanwhile, the sparse array is a simplified scheme, the number of the transducers can be reduced under the condition that the total aperture of the ultrasonic transducer is not changed, and the industrial implementation is facilitated. However, the acoustic signals acquired by the sparse array are incomplete, and images obtained by using a traditional reconstruction algorithm generate serious artifacts, so that the imaging quality is seriously reduced. Therefore, there is an urgent need to solve the problem of high-quality photoacoustic image reconstruction under sparse arrays. Deep learning techniques based on convolutional neural networks have been studied to recover high quality imaging results from blurred images, which are limited to a limited set of image data. The deep learning technology based on the recurrent neural network can predict the acoustic wave signals of missing array elements at other positions on the limited acoustic wave signals of the ultrasonic transducer to form complete annular array signals, and then a high-quality imaging result is recovered by utilizing a reconstruction algorithm.
Disclosure of Invention
The invention provides a photoacoustic imaging annular sparse array signal prediction method aiming at the condition that limited signal reconstruction images acquired by the existing photoacoustic imaging annular sparse array have artifacts and influence on imaging quality, so as to fill up limited signals of the sparse array.
The purpose of the invention is realized by the following technical scheme: the first aspect of the embodiment of the present invention provides a photoacoustic imaging annular sparse array signal prediction method, which specifically includes the following steps:
s1, normalizing the input signal and cutting off the signal of the preset initial stage length;
s2, dividing the processed input signal into a training data set, a label set of the training data set, a test data set and a label set of the test data set;
s3, inputting the training data set and the label set of the training data set into a convolution gating cyclic neural network for training to obtain a trained convolution gating cyclic neural network;
s4, inputting the test data set and the label set of the test data set into the convolution gate-controlled circular neural network obtained by training for verification on the dense array, and updating the convolution gate-controlled circular neural network;
and S5, acquiring signals in the sparse array, inputting the signals acquired in the sparse array into an updated convolution gating cyclic neural network for prediction, and obtaining signal data of ultrasonic transducer array elements at other positions of the sparse array.
Further, the ultrasonic transducer detects a sound pressure signal reflected by the tissue of the object to be measured, converts the sound pressure signal into an electrical signal, reads the electrical signal, and takes the electrical signal as an input signal; the input signals are N multiplied by M-shaped matrixes, M is the number of the ultrasonic transducer array elements, and N is the length of the signals collected by the single ultrasonic transducer array element.
Further, the length of the preset initial stage is 400-600, so that echo signals reflected by the ultrasonic transduction body and irrelevant to the target object are deleted.
Further, an annular ultrasonic transducer array containing S ultrasonic transducer array elements is used as a dense array, and each ultrasonic transducer is numbered as 1,2,3 … and 512 … S in sequence; forming an annular ultrasonic transducer array of S/4 ultrasonic transducer array elements in a mode that one ultrasonic transducer array element is reserved every other 4 ultrasonic transducer array elements, taking the annular ultrasonic transducer array containing the S/4 ultrasonic transducer array elements as a sparse array, and numbering each ultrasonic transducer as 1,5,9 … S-3 in sequence.
Further, the step S2 is specifically:
sequentially taking out signals of ultrasonic transducer array elements numbered as 1+4K from signals acquired by the ultrasonic transducer array elements numbered as 1 to serve as a training data set;
sequentially taking out ultrasonic transducer signals numbered as 2+4K from signals acquired by the ultrasonic transducer numbered as 2, and taking the ultrasonic transducer signals as a label set of a training data set;
sequentially taking out the ultrasonic transducer signals with the number of 3+4K from the signals collected by the ultrasonic transducer with the number of 3 as a test data set;
sequentially taking out ultrasonic transducer signals numbered as 4+4K from signals acquired by an ultrasonic transducer numbered as 4 to serve as a label set of a test data set; wherein K is 0,1,2 … 127 … S/4-1.
Furthermore, the convolution gating cyclic neural network consists of two layers of gating unit networks and a full connection layer; the number of units of each layer of the gate control unit network is equal to 128, the discarding rate of each layer is equal to 0.1, and the number of units of the last layer of the fully-connected layer is 1.
Further, in the process of training the convolutional gated cyclic neural network in step S3, the number of training iterations and the data amount of each batch are set by user, a root mean square back propagation algorithm optimizer is used, a Sigmoid function is used as an activation function, and an average absolute error is used as a loss function.
Further, the step S5 is specifically: collecting signals in a sparse array, standardizing the signals, and cutting off the signals of a preset initial stage length; inputting the signal into an updated convolution gating cyclic neural network for prediction; the prediction is carried out in three times, limited S/4 groups of array element data in the sparse array are used for the first time, and array element signals with the number of 2+4K are predicted; secondly, inputting the array element signals with the number of 2+4K obtained by prediction into a convolution gating cyclic neural network, and predicting the array element signals with the number of 3+ 4K; thirdly, inputting the array element signal with the number of 3+4K obtained by prediction into a convolution gating cyclic neural network, and predicting the array element signal with the number of 4+ 4K; wherein K is 0,1,2 … 127 … S/4-1; and obtaining signal data of the ultrasonic transducer array elements at other positions of the sparse array.
A second aspect of embodiments of the present invention provides an electronic device, comprising a memory and a processor, the memory being coupled to the processor; wherein the memory is used for storing program data, and the processor is used for executing the program data to realize the depth learning-based photoacoustic imaging annular sparse array signal prediction method.
A third aspect of embodiments of the present invention provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, implements the above-described depth learning-based photoacoustic imaging annular sparse array signal prediction method.
The invention has the beneficial effects that: the invention discloses a depth learning signal prediction method from an ultrasonic transducer signal end to a signal end, which does not need to use photoacoustic image data and provides a convenient way for signal filling of a sparse array and subsequent image reconstruction. Through WWCG-Net training prediction, the problem that limited signal reconstruction images acquired by the photoacoustic imaging annular sparse array have artifacts is solved, and imaging quality is improved.
Drawings
FIG. 1 is a schematic diagram of a dense array apparatus according to the present invention;
FIG. 2 is a schematic diagram of a sparse array device according to the present invention;
FIG. 3 is a flow chart of the method of the present invention;
FIG. 4 is a schematic diagram of a convolutional gated cyclic neural network structure from an acoustic signal of an ultrasonic transducer to an acoustic signal according to the present invention;
FIG. 5 is a diagram illustrating a model training loss value convergence process;
FIG. 6 is a graph of the comparison of the actual and predicted values of an ultrasound transducer signal numbered 401 in an example of the present invention;
FIG. 7 is a graph of sparse array signal strength in an example of the present invention;
FIG. 8 is a graph of the predicted dense array signal strength over WWCG-Net using the sparse array signal in an example of the present invention;
fig. 9 is a schematic structural diagram of a computer device according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the invention, as detailed in the appended claims.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
The photoacoustic imaging annular sparse array signal prediction method based on deep learning provided by the invention is further explained below by combining a specific example and the attached drawings. The features of the following examples and embodiments may be combined with each other without conflict.
The photoacoustic imaging experimental device has high requirements on the quality of the ultrasonic transducer array elements and higher construction cost. The sparse array may reduce the number of ultrasound transducer elements and effectively cover the target field of view. The data collected by the sparse array has low cost and quick response. However, the small number of array elements causes the photoacoustic imaging reconstruction result to have artifacts, and the imaging quality is affected. Through the Convolution Gate-controlled cyclic neural network (WWCG-Net) from the acoustic signal of the ultrasonic transducer to the acoustic signal, the signal data of the adjacent array elements and the array elements at other positions of the annular ultrasonic transducer can be predicted from the input sparse array photoacoustic signal data, and a complete photoacoustic signal is formed to obtain a high-quality photoacoustic imaging reconstruction result.
Taking an annular ultrasonic transducer array comprising S ultrasonic transducer array elements as a dense array, and numbering each ultrasonic transducer as 1,2,3 … and 512 … S in sequence; forming an S/4 array element annular ultrasonic transducer array in a mode that one array element is reserved at every 4 array elements, taking the S/4 array element annular ultrasonic transducer array as a sparse array, and numbering each ultrasonic transducer as 1,5,9 … S-3 in sequence.
In the embodiment of the invention, 512 ultrasonic transducer elements arranged in an annular ultrasonic transducer array are called as a dense array, each ultrasonic transducer is numbered as 1,2,3 … and 512, the system construction cost is high, and the data acquisition is slow. Thus, one quarter of the array elements of the ultrasound transducer, i.e. 128 ultrasound transducers, are retained, and a sparse array of 128 elements, hereinafter referred to as "sparse array", is formed with one retained every 4 elements, and each ultrasound transducer is numbered 1,5,9 … 509. Fig. 1 is a schematic diagram of a dense array, and fig. 2 is a schematic diagram of a sparse array.
Recording transducer signals collected by a dense array
The ultrasonic transducer signals collected by the sparse array are recorded as
As shown in fig. 3, the photoacoustic imaging sparse array signal prediction method based on WWCG-Net provided by the present invention includes the following steps:
s1, normalizing the input signal and truncating the signal for a predetermined start stage length.
First, data preprocessing is performed. The ultrasonic transducer detects a sound pressure signal reflected by a tissue of a measured object, converts the sound pressure signal into an electrical signal, reads the electrical signal, and takes the electrical signal as an input signal. Input signal
The array is an N multiplied by M matrix, wherein M is the number of ultrasonic transducer array elements in the array, and N is the length of a signal acquired by a single ultrasonic transducer array element. If the number of the array elements of the ultrasonic transducer is one, the input signal is in a vector form; if the number of the ultrasonic transducer array elements is multiple, the input signal is in a matrix form. Will input the signal
Carrying out standardization treatment, wherein the formula of the standardization treatment is as follows:
wherein,
is a signal after the normalization process and is,
is the average value of the average of the values,
is the standard deviation. Second, the length of the initial phase of the truncated signal ishThe signal(s) is (are) transmitted,hequal to any positive integer between 400 and 600, this portion of the signal reflects the acoustic signal reflected by the ultrasonic transducer machine itself, independent of the acoustic signal reflected by the target object. The signal processed in step S1 is still recorded as
The signal length becomes L, L = N-h.
S2, dividing the processed input signal into a training data set, a label set of the training data set, a test data set and a label set of the test data set.
Signal output from step S1
The method is divided into four parts, namely a training data set, a label set of the training data set, a test data set and a label set of the test data set. The training data set consists of the following methods: from the signal collected by the ultrasonic transducer numbered 1, ultrasonic transducer signals numbered 1+4K (K =0,1,2 … 127) are sequentially taken out, and a matrix having a shape of (128, L) is sequentially arranged. The label set of the training data set is formed by the following method: ultrasonic transducer signals numbered 2+4K (K =0,1,2 … 127) are sequentially taken out from the signal acquired by the ultrasonic transducer numbered 2, and are sequentially arranged into a matrix having a shape of (128, L). The test data set consists of the following methods: from the signal collected by the ultrasonic transducer numbered 3, the ultrasonic transducer signals numbered 3+4K (K =0,1,2 … 127) are sequentially taken out, and a matrix having a shape of (128, L) is sequentially arranged. The label set of the test data set is composed of the following methods: from the signal collected by the ultrasonic transducer numbered 4, the ultrasonic transducer signals numbered 4+4K (K =0,1,2 … 127) are sequentially taken out, and a matrix having a shape of (128, L) is sequentially arranged. The training data set and the test data set with the shapes of (128, L) are respectively expanded into one-dimensional vectors with the shapes of (m, 1), m =128 xL, and the same is carried out below. Continuously, the vector data with the shape of (m, 1) of the training data set and the test data set are unchanged, one dimension is added, and the shape of the vector data is changed into (m, 1, 1) to form a new training data set and a new test data set. And (3) respectively unfolding the training data label set and the test data label set with the shapes of (128, L) into one-dimensional vectors with the shapes of (m, 1) to form a new training data label set and a new test data label set.
And S3, inputting the training data set and the label set of the training data set into a convolution gating cyclic neural network for training to obtain the trained convolution gating cyclic neural network.
As shown in fig. 4, training data and a training data label set in a training data set are input into WWCG-Net for training, in this embodiment, the number of iterations epoch =130, the data amount of each batch is equal to 512, a Root Mean square back propagation (RMSprop) optimizer is used, an activation function is a Sigmoid function, and a loss function is an average Absolute Error (MAE). WWCG-Net uses two layers of gating cell networks and one layer of fully connected layers. The number of cells per layer of the gated cell network is equal to 128, the drop rate per layer is equal to 0.1, and the number of cells in the last fully connected layer is 1. Normally, the training loss value loss is kept stable and gradually reduced to converge for more than 60 rounds, and as shown in fig. 5, the model training is completed and the trained model is saved.
And S4, inputting the test data set and the label set of the test data set into the convolution gating cyclic neural network obtained by training for verification on the dense array, and updating the convolution gating cyclic neural network.
And testing data on a known dense array, taking a group of point light source experimental data as an example, wherein L is 1249, testing the trained model by using the test data, and outputting a WWCG-Net prediction result of the test data, wherein the prediction result is a one-dimensional vector with the shape of (m, 1), and the vector is restored into a matrix with the shape of (128, L). The prediction result of the model shows goodness of fit R2Values above 0.9 indicate better prediction results. As shown in fig. 6, the predicted data of the ultrasonic transducer, numbered 401, is compared with the corresponding real data. The dotted line represents the true value in the test data label set, the solid line represents the predicted value of WWCG-Net, the curves corresponding to the true value and the predicted value in the graph are almost overlapped, and the comparison result in the graph shows that the method has accurate prediction results on ultrasonic transducer signals at other positions in the array.
And S5, acquiring signals in the sparse array, inputting the signals into an updated convolution gating cyclic neural network for prediction, and obtaining signal data of the ultrasonic transducer array elements at other positions of the sparse array.
In the embodiment of the invention, in a sparse array, limited 128 groups of array elements are acquired, and data unknown to ultrasonic transducer elements at other positions in the sparse array is utilized.
Reading signals in the order of ultrasonic transducer numbering
The matrices having the shape of (128, L) are sequentially arranged, normalization processing is performed, and a signal having a signal start stage length of 500 is cut off. The normalized signal is expanded into a vector having a shape of (m, 1, 1), input to WWCG-Net for prediction, and finally the prediction result is restored into a matrix having a shape of (128, L). Prediction will be done in 3 steps, the first timeArray element signals numbered 2+4K (K =0,1,2 … 127) are predicted with a limited 128 sets of array element data in the sparse array. And inputting the array element signal with the number of 2+4K predicted by the invention into WWCG-Net for the second time, and predicting the array element signal with the number of 3+ 4K. And thirdly, inputting the array element signal with the number of 3+4K predicted by the invention into WWCG-Net to predict the array element signal with the number of 4+ 4K. The number of times of prediction and the number of array elements to be filled can be changed according to the actual situation. In a sparse array, a dense set of 512 ultrasonic transducer element signals can be obtained by the invention. Taking a group of simulated sparse array point light source data as an example, as shown in fig. 7, the signals recovered by using the sparse array signal prediction method of the present invention are denser, the blank area of the sparse array signal is effectively filled, and the predicted result is shown in fig. 8.
The annular array verified by the embodiment of the invention can be placed with 512 ultrasonic transducer array elements at the maximum, and the condition of the sparse array is that no more than 128 ultrasonic transducer array elements are placed at equal intervals. The invention provides a signal prediction method from the signal end of an ultrasonic transducer to the signal end, and provides a convenient path for filling signals of a sparse array and reconstructing subsequent images.
In summary, the invention provides a method for predicting a deep learning signal from an ultrasonic transducer signal end to a signal end, a time sequence signal is used for prediction, photoacoustic image data is not needed, and a convenient path is provided for signal filling of a sparse array and subsequent image reconstruction. Through WWCG-Net training prediction, the problem that limited signal reconstruction images acquired by the photoacoustic imaging annular sparse array have artifacts is solved, and imaging quality is improved.
Corresponding to the embodiment of the photoacoustic imaging annular sparse array signal prediction method based on the deep learning, the invention also provides an embodiment of a photoacoustic imaging annular sparse array signal prediction device based on the deep learning.
Referring to fig. 9, a photoacoustic imaging circular sparse array signal prediction apparatus provided in an embodiment of the present invention includes one or more processors, and is configured to implement the photoacoustic imaging circular sparse array signal prediction method in the foregoing embodiment.
The embodiment of the photoacoustic imaging annular sparse array signal prediction apparatus of the present invention can be applied to any device with data processing capability, such as a computer or other devices or apparatuses. The device embodiments may be implemented by software, or by hardware, or by a combination of hardware and software. The software implementation is taken as an example, and as a logical device, the device is formed by reading corresponding computer program instructions in the nonvolatile memory into the memory for running through the processor of any device with data processing capability. In terms of hardware, as shown in fig. 9, a hardware structure diagram of any device with data processing capability where the photoacoustic imaging circular sparse array signal prediction apparatus of the present invention is located is shown, except for the processor, the memory, the network interface, and the nonvolatile memory shown in fig. 9, in the embodiment, any device with data processing capability where the apparatus is located may generally include other hardware according to the actual function of the any device with data processing capability, which is not described again.
The implementation process of the functions and actions of each unit in the above device is specifically described in the implementation process of the corresponding step in the above method, and is not described herein again.
For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of the invention. One of ordinary skill in the art can understand and implement it without inventive effort.
An embodiment of the present invention also provides a computer-readable storage medium on which a program is stored, which, when executed by a processor, implements the photoacoustic imaging annular sparse array signal prediction method in the above-described embodiments.
The computer readable storage medium may be an internal storage unit, such as a hard disk or a memory, of any data processing capability device described in any of the foregoing embodiments. The computer readable storage medium can be any device with data processing capability, such as a plug-in hard disk, a Smart Media Card (SMC), an SD Card, a Flash memory Card (Flash Card), etc. provided on the device. Further, the computer readable storage medium may include both an internal storage unit and an external storage device of any data processing capable device. The computer-readable storage medium is used for storing the computer program and other programs and data required by the arbitrary data processing-capable device, and may also be used for temporarily storing data that has been output or is to be output.
The above embodiments are only used for illustrating the design idea and features of the present invention, and the purpose of the present invention is to enable those skilled in the art to understand the content of the present invention and implement the present invention accordingly, and the protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes and modifications made in accordance with the principles and concepts disclosed herein are intended to be included within the scope of the present invention.
Claims (10)
1. A photoacoustic imaging annular sparse array signal prediction method is characterized by specifically comprising the following steps of:
s1, normalizing the input signal and cutting off the signal of the preset initial stage length;
s2, dividing the processed input signal into a training data set, a label set of the training data set, a test data set and a label set of the test data set;
s3, inputting the training data set and the label set of the training data set into a convolution gating cyclic neural network for training to obtain a trained convolution gating cyclic neural network;
s4, inputting the test data set and the label set of the test data set into the convolution gate-controlled circular neural network obtained by training for verification on the dense array, and updating the convolution gate-controlled circular neural network;
and S5, acquiring signals in the sparse array, inputting the signals acquired in the sparse array into an updated convolution gating cyclic neural network for prediction, and obtaining signal data of ultrasonic transducer array elements at other positions of the sparse array.
2. The photoacoustic imaging annular sparse array signal prediction method of claim 1, wherein the ultrasonic transducer detects a sound pressure signal reflected by a tissue of the object to be measured, converts the sound pressure signal into an electrical signal, reads the electrical signal, and takes the electrical signal as an input signal; the input signal is an N multiplied by M matrix, M is the number of ultrasonic transducer array elements in the array, and N is the signal length acquired by a single ultrasonic transducer array element.
3. The method for predicting the photoacoustic imaging annular sparse array signal according to claim 1, wherein the preset initial stage length is 400-600 to delete the echo signal reflected by the ultrasonic transducer independent of the target object.
4. The photoacoustic imaging annular sparse array signal prediction method of claim 1, wherein an annular ultrasound transducer array comprising S ultrasound transducer array elements is taken as a dense array, and each ultrasound transducer is numbered sequentially as 1,2,3 …,512 … S; forming an annular ultrasonic transducer array of S/4 ultrasonic transducer array elements in a mode that one ultrasonic transducer array element is reserved every other 4 ultrasonic transducer array elements, taking the annular ultrasonic transducer array containing the S/4 ultrasonic transducer array elements as a sparse array, and numbering each ultrasonic transducer as 1,5,9 … S-3 in sequence.
5. The photoacoustic imaging annular sparse array signal prediction method of claim 4, wherein the step S2 is specifically:
sequentially taking out signals of ultrasonic transducer array elements numbered as 1+4K from signals acquired by the ultrasonic transducer array elements numbered as 1 to serve as a training data set;
sequentially taking out ultrasonic transducer signals numbered as 2+4K from signals acquired by the ultrasonic transducer numbered as 2, and taking the ultrasonic transducer signals as a label set of a training data set;
sequentially taking out the ultrasonic transducer signals with the number of 3+4K from the signals collected by the ultrasonic transducer with the number of 3 as a test data set;
sequentially taking out ultrasonic transducer signals numbered as 4+4K from signals acquired by an ultrasonic transducer numbered as 4 to serve as a label set of a test data set; wherein K is 0,1,2 … 127 … S/4-1.
6. The photoacoustic imaging annular sparse array signal prediction method of claim 1, wherein the convolutional gated cyclic neural network consists of two layers of gated cell networks and one layer of fully connected layer; the number of units of each layer of the gate control unit network is equal to 128, the discarding rate of each layer is equal to 0.1, and the number of units of the last layer of the fully-connected layer is 1.
7. The method for predicting the photoacoustic imaging annular sparse array signal according to claim 1, wherein in the step S3 of training the convolutional gated cyclic neural network, the number of training iterations and the data amount of each batch are set by user, a root-mean-square back-propagation algorithm optimizer is used, a Sigmoid function is used as an activation function, and an average absolute error is used as a loss function.
8. The photoacoustic imaging annular sparse array signal prediction method of claim 4, wherein the step S5 is specifically: collecting signals in a sparse array, standardizing the signals, and cutting off the signals of a preset initial stage length; inputting the signal into an updated convolution gating cyclic neural network for prediction; the prediction is carried out for 3 times, limited S/4 array element data in the sparse array is used for the first time, and array element signals with the number of 2+4K are predicted; secondly, inputting the array element signals with the number of 2+4K obtained by prediction into a convolution gating cyclic neural network, and predicting the array element signals with the number of 3+ 4K; thirdly, inputting the array element signal with the number of 3+4K obtained by prediction into a convolution gating cyclic neural network, and predicting the array element signal with the number of 4+ 4K; wherein K is 0,1,2 … 127 … S/4-1; and obtaining signal data of the ultrasonic transducer array elements at other positions of the sparse array.
9. An electronic device comprising a memory and a processor, wherein the memory is coupled to the processor; wherein the memory is used for storing program data, and the processor is used for executing the program data to realize the photoacoustic imaging annular sparse array signal prediction method of any one of claims 1 to 8.
10. A computer-readable storage medium on which a computer program is stored, which when executed by a processor implements the photoacoustic imaging circular sparse array signal prediction method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210450980.XA CN114548191B (en) | 2022-04-27 | 2022-04-27 | Photoacoustic imaging annular sparse array signal prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210450980.XA CN114548191B (en) | 2022-04-27 | 2022-04-27 | Photoacoustic imaging annular sparse array signal prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114548191A true CN114548191A (en) | 2022-05-27 |
| CN114548191B CN114548191B (en) | 2022-08-05 |
Family
ID=81667323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210450980.XA Active CN114548191B (en) | 2022-04-27 | 2022-04-27 | Photoacoustic imaging annular sparse array signal prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114548191B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114859360A (en) * | 2022-07-08 | 2022-08-05 | 之江实验室 | Ultrasonic synthetic aperture imaging method and device introducing machine learning |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102727259A (en) * | 2012-07-26 | 2012-10-17 | 中国科学院自动化研究所 | Photoacoustic tomography device and method based on limited-angle scanning |
| US20140036091A1 (en) * | 2011-11-02 | 2014-02-06 | Seno Medical Instruments, Inc. | Interframe energy normalization in an optoacoustic imaging system |
| CN109242400A (en) * | 2018-11-02 | 2019-01-18 | 南京信息工程大学 | A kind of logistics express delivery odd numbers recognition methods based on convolution gating cycle neural network |
| US20200311932A1 (en) * | 2019-03-28 | 2020-10-01 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Synthetic Medical Image Generation |
| CN111956180A (en) * | 2019-05-20 | 2020-11-20 | 华北电力大学（保定） | Method for reconstructing photoacoustic endoscopic tomography image |
| CN112683808A (en) * | 2020-12-15 | 2021-04-20 | 电子科技大学 | Multi-component device based on photoacoustic spectrum and signal processing method |
| US20210272339A1 (en) * | 2018-07-06 | 2021-09-02 | Mayo Foundation For Medical Education And Research | Systems and Methods for Generating and Estimating Unknown and Unacquired Ultrasound Data |
| CN113607068A (en) * | 2021-07-19 | 2021-11-05 | 华中科技大学 | Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics |
| CN114332283A (en) * | 2021-12-31 | 2022-04-12 | 之江实验室 | Training method based on double-domain neural network and photoacoustic image reconstruction method |
-
2022
- 2022-04-27 CN CN202210450980.XA patent/CN114548191B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140036091A1 (en) * | 2011-11-02 | 2014-02-06 | Seno Medical Instruments, Inc. | Interframe energy normalization in an optoacoustic imaging system |
| CN102727259A (en) * | 2012-07-26 | 2012-10-17 | 中国科学院自动化研究所 | Photoacoustic tomography device and method based on limited-angle scanning |
| US20210272339A1 (en) * | 2018-07-06 | 2021-09-02 | Mayo Foundation For Medical Education And Research | Systems and Methods for Generating and Estimating Unknown and Unacquired Ultrasound Data |
| CN109242400A (en) * | 2018-11-02 | 2019-01-18 | 南京信息工程大学 | A kind of logistics express delivery odd numbers recognition methods based on convolution gating cycle neural network |
| US20200311932A1 (en) * | 2019-03-28 | 2020-10-01 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Synthetic Medical Image Generation |
| CN111956180A (en) * | 2019-05-20 | 2020-11-20 | 华北电力大学（保定） | Method for reconstructing photoacoustic endoscopic tomography image |
| CN112683808A (en) * | 2020-12-15 | 2021-04-20 | 电子科技大学 | Multi-component device based on photoacoustic spectrum and signal processing method |
| CN113607068A (en) * | 2021-07-19 | 2021-11-05 | 华中科技大学 | Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics |
| CN114332283A (en) * | 2021-12-31 | 2022-04-12 | 之江实验室 | Training method based on double-domain neural network and photoacoustic image reconstruction method |
Non-Patent Citations (5)
| Title |
|---|
| SIMON ARRIDGE ET AL: "Accelerated high-resolution photoacoustic tomography via compressed sensing", 《PHYSICS IN MEDICINE & BIOLOGY》 * |
| TANJI KOICHI ET AL: "Advanced model-based reconstruction algorithm for practical three-dimensional photoacoustic imaging", 《CONFERENCE ON PHOTONS PLUS ULTRASOUND - IMAGING AND SENSING》 * |
| 刘子超: "基于深度学习的肿瘤光声图像重建方法研究", 《中国优秀硕士学位论文全文数据库》 * |
| 赵明珠: "基于多频段融合的光声无创血糖检测系统设计", 《中国优秀硕士学位论文全文数据库》 * |
| 闫向阳: "采用稀疏测量数据的有限角度扫描内窥式光声图像重建的研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114859360A (en) * | 2022-07-08 | 2022-08-05 | 之江实验室 | Ultrasonic synthetic aperture imaging method and device introducing machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114548191B (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Alvarez et al. | Learning the number of neurons in deep networks | |
| CN110074813B (en) | Ultrasonic image reconstruction method and system | |
| CN111383741B (en) | Method, device and equipment for establishing medical imaging model and storage medium | |
| CN111445390A (en) | Wide residual attention-based three-dimensional medical image super-resolution reconstruction method | |
| CN112529146B (en) | Neural network model training method and device | |
| CN112329609A (en) | Feature fusion transfer learning arrhythmia classification system based on 2D heart beat | |
| CN114548191B (en) | Photoacoustic imaging annular sparse array signal prediction method and device | |
| CN111680755B (en) | Medical image recognition model construction and medical image recognition method, device, medium and terminal | |
| CN111383742A (en) | Method, device, equipment and storage medium for establishing medical imaging model | |
| CN111915691A (en) | Image processing system, method, terminal and medium based on neural network | |
| CN112614229A (en) | Face reconstruction method and device based on deep learning | |
| Sharifzadeh et al. | Ultrasound domain adaptation using frequency domain analysis | |
| CN110246200B (en) | Magnetic resonance cardiac cine imaging method and device and magnetic resonance scanner | |
| CN114972570A (en) | Image reconstruction method and device | |
| CN110570375A (en) | image processing method, image processing device, electronic device and storage medium | |
| JPH04242865A (en) | Data processing device and method selecting data word included in dictionary | |
| CN118039124A (en) | Object analysis method, device, computer equipment and storage medium | |
| CN116402995A (en) | Lightweight neural network-based ancient architecture point cloud semantic segmentation method and system | |
| CN113947547B (en) | Monte Carlo rendering graph noise reduction method based on multi-scale kernel prediction convolutional neural network | |
| CN113362292B (en) | Bone age assessment method and system based on programmable logic gate array | |
| Li et al. | Latent source mining of fMRI data via deep belief network | |
| CN112669450A (en) | Human body model construction method and personalized human body model construction method | |
| CN116563673B (en) | Smoke training data generation method and device and computer equipment | |
| CN115409952A (en) | Tuberculous meningoencephalitis reconstruction system, method and memory | |
| CN113409326B (en) | Image segmentation method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |