Method and device for predicting photoacoustic imaging array signals for limited field of view sampling

Abstract

The present invention discloses a method and device for predicting photoacoustic imaging array signals for limited field of view sampling. The method includes: detecting the acoustic pressure signals reflected by the tissue of the object to be measured through ultrasonic transducers in a photoacoustic imaging array, converting the acoustic pressure signals into electrical signals, and performing stationarity tests on the electrical signals; selecting the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combining the truncation number and the trailing number pairwise to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion; determining an autoregressive moving average model according to the order, and then weighting it with an LSTM network to obtain a photoacoustic imaging array signal prediction model; inputting the electrical signals obtained after the stationarity test into the photoacoustic imaging array signal prediction model for prediction, and ending the prediction when the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, so as to obtain the predicted photoacoustic signals.

Description

Technical Field

[0001] The present invention belongs to the fields of artificial intelligence and biomedical engineering, and particularly relates to a method and device for predicting photoacoustic imaging array signals for limited field of view sampling. Background Art

[0002] Photoacoustic imaging is a new non-invasive and non-ionizing biomedical imaging method that combines the advantages of high contrast of optics and high penetration of acoustics inside biological tissues, and has great clinical application prospects in detecting blood oxygen content and malignant tumors near blood vessels. The number, arrangement method, and arrangement density of ultrasonic transducers in the array have a great impact on the results of photoacoustic imaging. High-quality photoacoustic imaging results require a high-density ultrasonic transducer array element and a large-range spatial arrangement to cover the complete detection perspective. However, limited by the geometric shape of the detected object and the imaging space, a high-density ultrasonic transducer array with full-field coverage cannot be placed around the object to be measured, and such an array is large in volume, high in cost, and not easy to be used clinically on a large scale. A small photoacoustic imaging array contains a small number of ultrasonic transducers. It can speed up the sampling speed under a limited sampling field of view by arranging the ultrasonic transducers in a specific manner, but there are defects such as image reconstruction distortion and interference of artifacts. With the development of artificial intelligence algorithms, machine learning algorithms can reconstruct high-quality photoacoustic imaging under limited sampling conditions, but a large amount of existing image data sets are required as training data, the training cost is relatively high, and the photoacoustic imaging reconstruction efficiency is relatively low.

[0003] Therefore, there is an urgent need to propose a method for predicting photoacoustic imaging array signals to obtain high-quality signal reconstruction results under the condition of a limited field of view. Summary of the Invention

[0004] In view of the deficiencies of the prior art, a method and device for predicting photoacoustic imaging array signals for limited field of view sampling are proposed.

[0005] The present invention is implemented through the following technical solutions:

[0006] According to the first aspect of the embodiments of the present invention, a method for predicting photoacoustic imaging array signals for limited field of view sampling is provided, and the method includes:

[0007] Detect the acoustic pressure signal reflected by the tissue of the object to be measured through the ultrasonic transducers in the photoacoustic imaging array, convert the acoustic pressure signal into an electrical signal, and perform a stationarity test on the electrical signal;

[0008] Select the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combine the truncation number and the trailing number in pairs to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion;

[0009] Determine the autoregressive moving average model according to the order, and then weight it with the LSTM network to construct a photoacoustic imaging array signal prediction model;

[0010] Input the electrical signal obtained after the stationarity test into the photoacoustic imaging array signal prediction model for prediction. When the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, end the prediction to obtain the predicted photoacoustic signals.

[0011] According to the second aspect of the embodiments of the present invention, there is provided a photoacoustic imaging array signal prediction device for limited field of view sampling, including one or more processors for the above-mentioned photoacoustic imaging array signal prediction method for limited field of view sampling.

[0012] According to the third aspect of the embodiments of the present invention, there is provided a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it is used for the above-mentioned photoacoustic imaging array signal prediction method for limited field of view sampling.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a photoacoustic imaging array signal prediction method and device for limited field of view sampling. Aiming at the situation of image reconstruction distortion and artifact interference in limited field of view sampling of a small photoacoustic imaging array, the truncation number and the trailing number are selected through the autocorrelation function and the partial autocorrelation function, and then the order of the photoacoustic imaging array signal prediction model is determined based on the Bayesian information criterion. According to the order, a photoacoustic imaging array signal prediction model is constructed based on the autoregressive moving average model and the LSTM network to predict the photoacoustic signals. By combining the predicted signals and the collected signals, an accurate target object imaging within the region to be measured is reconstructed. Description of the Drawings

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a schematic diagram of limited sampling of a small photoacoustic imaging array provided by the embodiments of the present invention;

[0016] Figure 2 It is a flowchart of the method of the present invention;

[0017] Figure 3 It is a flow block diagram of the method of the present invention;

[0018] Figure 4It is a schematic diagram for testing the stability of photoacoustic signals mentioned in the embodiments of the present invention;

[0019] Figure 5 It is a schematic diagram for determining the order of the photoacoustic imaging array signal prediction model PSPM in the embodiments of the present invention;

[0020] Figure 6 It is a schematic diagram of the LSTM network structure proposed by the present invention;

[0021] Figure 7 It is a comparison result diagram of the signal prediction value and the true value of the photoacoustic imaging array signal prediction model PSPM proposed by the present invention;

[0022] Figure 8 It is a photoacoustic image reconstruction result diagram of a simulated point light source under limited sampling of a quarter-ring array in the embodiments of the present invention;

[0023] Figure 9 It is a reconstruction result diagram of a simulated point light source under limited sampling of a quarter-ring array using the photoacoustic imaging array signal prediction model PSPM in the embodiments of the present invention;

[0024] Figure 10 It is a schematic diagram of an electronic device provided in the embodiments of the present invention. Detailed implementation manners

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] It should be noted that, without conflict, the features in the following embodiments and implementation manners can be combined with each other.

[0027] The full-ring photoacoustic imaging device has a high-density and high-precision ultrasonic transducer array, which has a high development cost, a large volume, and is not easy to be used clinically. The handheld small photoacoustic imaging array is usually a linear array or a quarter-ring array. This kind of array has the advantages of low cost and easy use. However, due to the low spatial element density, there are deviations in the imaging results, and accurate full-space imaging cannot be formed under a limited detection field of view.

[0028] The present invention aims at Figure 1For the photoacoustic imaging array shown, which is non-fully annular and cannot cover a sufficient imaging area, a signal prediction method for a photoacoustic imaging array for limited field of view sampling is provided. By combining the Autoregressive Moving Average model (ARMA) and the Long Short-Term Memory Networks (LSTM), the received signals of ultrasonic transducers in the areas not covered by the field of view in the array are predicted from the photoacoustic signals sampled limitedly, so as to increase the array signal density, and then the back-projection reconstruction algorithm is used to reconstruct accurate and high-quality photoacoustic imaging results that can include the full field of view.

[0029] As Figure 2 and Figure 3 shown, a signal prediction method for a photoacoustic imaging array for limited field of view sampling proposed in an embodiment of the present invention may include the following steps:

[0030] Step S1, detecting the acoustic pressure signal reflected by the tissue of the object to be measured through the ultrasonic transducers in the photoacoustic imaging array, converting the acoustic pressure signal into an electrical signal, and performing a stationarity test.

[0031] The electrical signal is represented as an M×N matrix, where N is the length of the signal collected by a single ultrasonic transducer, and M is the number of ultrasonic transducers. Each row of this matrix, that is, a vector of size (1, N), is denoted as the signal s k (t), k = 1, 2... M. After expanding this matrix, that is, a vector of size (1, M×N) is denoted as the signal S t , and the signal S t is used as the input signal of the photoacoustic imaging array signal prediction model (PSPM) proposed in the present invention.

[0032] Performing the stationarity test includes: drawing the waveform diagram of the section from t = N / 4 to N*3 / 4 of S t to observe whether the input signal fluctuates above and below the mean value. Secondly, perform the Augmented Dickey-Fuller test (ADF) and the KPSS (Kwiatkowski-Phillips-Schmidt-Shin Test) on the input signal S t . When the ADF value of the test result is equal to 1 and the KPSS value is equal to 0, the test is passed, and this input signal is a stationary time series signal. If the ADF value of the test result is not equal to 0 or the KPSS value is not equal to 1, perform differencing processing on the input signal S t and then perform the test until the stationarity test is passed.

[0033] Step S2: Select the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combine the truncation number and the trailing number pairwise to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion.

[0034] When implementing Step S2, too small an order will cause the model to not accurately reflect the signal trend, and too large an order will cause the model to overfit. Specifically, in this example, the truncation number and the trailing number are selected by plotting the autocorrelation (ACF) lag coefficient graph and the partial autocorrelation (PACF) lag coefficient graph. The truncation number and the trailing number are combined pairwise, and the Bayesian information criterion (BIC) of each combination is calculated to obtain the Bayesian information criterion (BIC) matrix. In the output BIC matrix, find the position of the smallest element in the BIC matrix, and record the position of this smallest element as the a-th row and the b-th column. Then, the order p of the photoacoustic imaging array signal prediction model is p = a - 1, and q = b - 1, so as to determine the optimal order of the photoacoustic imaging array signal prediction model.

[0035] Among them, the photoacoustic imaging array signal prediction model can be expressed as:

[0036] s k (t) = αm k (t) + βn k (t) = α(γ 0 + γ 1 S t-1 + … + γ a S t-p + ε t + c 1 ε t-1 + … + c b ε t-q ) + βn k (t)

[0037] Among them, s k (t) is the prediction result of the photoacoustic imaging array signal prediction model for the signal collected by the k-th ultrasonic transducer, m k (t) is the prediction result of the autoregressive moving average model for the signal collected by the k-th ultrasonic transducer, n k (t) is the prediction result of the LSTM network for the signal collected by the k-th ultrasonic transducer, α and β are weight coefficients, and α + β = 1.

[0038] γ i is the autoregressive coefficient, i = 0, 1…p. S t-p is the amplitude of the input signal S t at time t - p. ε t-j is the time series of the error term, c j is the moving average coefficient, j = 1…q.

[0039] S3. Determine the autoregressive moving average model according to the order obtained in step S2, and then weight it with the LSTM network to obtain the photoacoustic imaging array signal prediction model PSPM.

[0040] S4. Input the information obtained after the stationarity test in step S1 into the photoacoustic imaging array signal prediction model for prediction. When the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, the prediction ends, and the predicted photoacoustic signals are output.

[0041] Embodiment 1

[0042] Take the 1 / 4 circular ultrasonic transducer array containing K ultrasonic transducer elements as the small photoacoustic imaging array under limited field of view sampling in this embodiment. In this embodiment, K = 16, the ultrasonic transducers are numbered 1, 2... K, and the signal length collected by a single ultrasonic transducer is 2000.

[0043] As Figure 2 and Figure 3 shown, a method for predicting photoacoustic imaging array signals for limited field of view sampling proposed in an embodiment of the present invention may include the following steps:

[0044] Step S1. The ultrasonic transducers in the photoacoustic imaging array detect the sound pressure signal reflected by the tissue of the object to be measured, convert the sound pressure signal into an electrical signal, read the electrical signal, use the electrical signal as the input signal, and perform a stationarity test.

[0045] This electrical signal is an M×N matrix, where N is the signal length collected by a single ultrasonic transducer, and M is the number of ultrasonic transducers. Each row of this matrix, that is, a vector of size (1, N), is regarded as the photoacoustic signal collected by a single ultrasonic transducer in the array, denoted as s k (t), i = 1, 2... M. After expanding this matrix, that is, a vector of size (1, M×N) is denoted as signal S t , and signal S t is used as the input signal of the photoacoustic imaging array signal prediction model (PSPM) proposed by the present invention.

[0046] In this embodiment, taking the data of the array imaging a hair as an example, draw the waveform diagram of the part of the input signal S t from t = 500 to 1500, as Figure 4As shown, observe whether the input signal fluctuates above and below the mean. Secondly, perform ADF test and KPSS test on the input signal. If the ADF value of the test result is equal to 1 and the KPSS value is equal to 0, then the test is passed, and this input signal is a stationary time series signal. If the ADF value of the test result is not equal to 0 or the KPSS value is not equal to 1, perform differencing on the input signal and then conduct the test until the test is passed.

[0047] Step S2: Select the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combine the truncation number and the trailing number in pairs to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion.

[0048] Calculate the order of the photoacoustic imaging array signal prediction model using the Bayesian information criterion (BIC), and draw the autocorrelation and partial correlation lag coefficient diagrams. The order of the photoacoustic imaging array signal prediction model cannot be selected only through the schematic diagrams of ACF and PACF. It is also necessary to further determine the optimal PSPM model order by fitting several models with different lag coefficients. Too small an order will cause the model to not accurately reflect the signal trend, and too large an order will cause the model to overfit. It can be known from Figure 5 that combinations of p = 0, 1, 2, 3 and q = 0, 1 are selected, a total of 8 combinations, and calculate the BIC of each combination to determine the optimal order of the photoacoustic imaging array signal prediction PSPM model.

[0049] The calculation results of the Bayesian information criterion matrix are as follows:

[0050]

[0051] In the output BIC matrix, the minimum value is in the second row and the second column. Therefore, the order of the PSPM model is p = 1 and q = 1. The photoacoustic imaging array signal prediction model PSPM can be expressed as:

[0052] s k (t)=αm k (t)+βn k (t)=α(γ 0 +γ 1 S t-1 +ε t +cε t-1 )+βn k (t)

[0053] Among them, s k (t) is the prediction result of the photoacoustic imaging array signal prediction model, m k (t) is the prediction result of the autoregressive moving average model, n k (t) is the prediction result of the LSTM network, and α and β are weights, with α + β = 1.

[0054] γ j is the autoregressive coefficient, j = 0, 1. S t-1 is the input signal s t at the amplitude at time t - 1. c is the moving average coefficient. ε t is the time series of the error term. In this embodiment, α = 0.2, β = 0.8.

[0055] S3. Determine the autoregressive moving average model according to the order obtained in step S2, and then weight it with the LSTM network to construct a photoacoustic imaging array signal prediction model.

[0056] S4. Use the photoacoustic imaging array signal prediction model PSPM to predict the input photoacoustic signal: input the electrical signal obtained after the stationarity test into the photoacoustic imaging array signal prediction model for prediction, and end the prediction when the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, and obtain the predicted photoacoustic signal. Combine the predicted photoacoustic signal and the original acquisition signal to expand the number of photoacoustic signals under limited sampling and form the photoacoustic signal under full - field sampling.

[0057] It should be noted that the signal length predicted by the photoacoustic imaging array signal prediction model PSPM each time is the same as the signal length collected by a single transducer in the input signal S t in.

[0058] First, perform L2 - norm normalization on the input photoacoustic signal S t to obtain the signal S t * ; The formula for L2 - norm normalization is:

[0059]

[0060] The prediction of the photoacoustic imaging array signal prediction model is divided into two parts:

[0061] The first part is the autoregressive moving average model prediction. Input the normalized photoacoustic signal S t * , in this embodiment, the first predicted signal is the received signal of the ultrasonic transducer numbered K = 17, and the predicted value of the autoregressive moving average model is m 17 (t).

[0062] The second part is the LSTM network prediction. The method also includes the training part of the LSTM network. Normalize the signal S t …Divide it into a training data set and a test data set in the ratio of 8:2. Input the training data into the LSTM network for training. The brief structure of the LSTM network for the photoacoustic imaging array signal prediction model PSPM model is as follows Figure 6 shown. In this embodiment, the number of iterations epoch = 30 times, the amount of data taken per batch is equal to 2048, the Root Mean Square prop (RMSprop) optimizer is used, the activation function is the Sigmoid function, and the loss function is the Mean Absolute Error (MAE). The LSTM network uses three layers of long short-term memory networks, a pooling layer and a fully connected layer. The number of units in the first layer of the long short-term memory network is equal to 128, and the number of units in the second and third layers of the long short-term memory network is equal to 64. The dropout rate of each layer is equal to 0.1, and the number of units in the last fully connected layer is 1. Usually, when the training loss value loss remains stable and gradually decreases and converges after more than 20 rounds, the model training is completed, and the trained model is saved. Use the trained LSTM network to predict the input signal S t * as the input of the LSTM network. The predicted value of the LSTM network is n 17 (t).

[0063] Then the predicted value of the signal received by the ultrasonic transducer numbered 17 by the photoacoustic imaging array signal prediction model PSPM is: s 17 (t) = 0.2m 17 (t) + 0.8n 17 (t). Connect the predicted result s 17 (t) signal to the tail of S t * to form a new input signal, and input it into the photoacoustic imaging array signal prediction model PSPM again for prediction to obtain the predicted signal s 18 (t), and so on.

[0064] Figure 7 Shows the comparison between the predicted value and the true value of s 17 (t) by the PSPM model of the present invention. The results show that the PSPM model can accurately predict the photoacoustic signals received by ultrasonic transducers outside the limited field of view.

[0065] Preset the number N of ultrasonic transducers in the photoacoustic imaging array. When the photoacoustic signal output by the photoacoustic imaging array signal prediction model is s N (t), end the prediction to obtain a combined signal including the actually received signal and the predicted signal: S t * =(s 1 (t), s2 (t),…,s k (t),s k+1 (t0,…,s N (t)).

[0066] The number of preset ultrasonic transducers verified in this embodiment is 64. That is, when the photoacoustic imaging array signal prediction model outputs the predicted signal s 64 (t), stop the prediction and output the combined signal containing the actually received signal and the predicted signal: S t * =(s 1 (t), s 2 (t)…s 64 (t)).

[0067] Finally, use the back-projection reconstruction algorithm for image reconstruction to obtain a high-quality and accurate photoacoustic reconstruction image of the array under a limited detection field of view.

[0068] As Figure 8 shown, it is the result of imaging a combination of a carbon rod and a hair in a 1 / 4 circular array containing 16 ultrasonic transducers in this embodiment. Under the 1 / 4 circular ultrasonic transducer array, due to the limited field of view, only limited sampling can be performed, and the imaging result obtained from this limited sampling signal cannot accurately reflect the spatial position and accurate structure of the target to be measured. Figure 9 It shows the result of image reconstruction by outputting the photoacoustic signals of 64 ultrasonic transducers after predicting the photoacoustic signals of limited sampling using the photoacoustic imaging array signal prediction model PSPM of the present invention. The imaging result has been greatly improved and can accurately reflect the spatial position relationship and accurate structure of the combination of the hair and the carbon rod.

[0069] In summary, the present invention is a weighted machine learning array photoacoustic signal prediction method that combines an autoregressive moving average model and a long short-term memory network. It can predict photoacoustic signals in the detection field of view area that cannot be covered by the array from the photoacoustic signals obtained by the ultrasonic transducer under the condition of limited field of view sampling, increase the array signal density, and provide a convenient way for array signal prediction and subsequent image reconstruction under limited detection field of view. Different from the deep learning methods that only require photoacoustic image data, the photoacoustic signal prediction method provided by the present invention does not require an image data set and does not require large-scale model training. It can efficiently and low-costly solve the problems of artifacts in image reconstruction and missing imaging targets caused by signal sampling in photoacoustic imaging under limited field of view, and improve the image reconstruction quality of photoacoustic imaging under limited field of view sampling. At the same time, the photoacoustic imaging array signal prediction model constructed by the present invention has a wide application range and low computational complexity, does not depend on external training data, and adjusts the fitting degree of the photoacoustic imaging array signal prediction model through weights to increase the application range of the photoacoustic imaging array signal prediction model for different imaging targets.

[0070] Corresponding to the foregoing embodiments of the photoacoustic imaging array signal prediction method for limited field of view sampling, the present invention also provides an embodiment of a photoacoustic imaging array signal prediction device for limited field of view sampling.

[0071] See Figure 10 , an embodiment of a photoacoustic imaging array signal prediction device for limited field of view sampling provided by an embodiment of the present invention includes one or more processors for implementing the photoacoustic imaging array signal prediction method for limited field of view sampling in the foregoing embodiments.

[0072] The embodiment of the photoacoustic imaging array signal prediction device of the present invention for limited field of view sampling can be applied to any device with data processing capabilities, and the any device with data processing capabilities can be a device or apparatus such as a computer. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a logically meaningful device, it is formed by the processor of any device with data processing capabilities where it is located reading the corresponding computer program instructions in the non-volatile memory into the memory for operation. From the hardware level, as Figure 10 shown, it is a hardware structure diagram of any device with data processing capabilities where the photoacoustic imaging array signal prediction device of the present invention for limited field of view sampling is located. In addition to the Figure 10 shown processor, memory, network interface, and non-volatile memory, any device with data processing capabilities where the device in the embodiment is located usually also includes other hardware according to the actual functions of the any device with data processing capabilities, which will not be elaborated here.

[0073] For the implementation processes of the functions and roles of each unit in the above device, please refer to the implementation processes of the corresponding steps in the above method for details, which will not be elaborated here.

[0074] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial descriptions of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0075] The embodiments of the present invention also provide a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the method for predicting photoacoustic imaging array signals for limited field of view sampling in the above embodiments is implemented.

[0076] The computer-readable storage medium can be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or a memory. The computer-readable storage medium can also be any device with data processing capabilities, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium can also include both an internal storage unit of any device with data processing capabilities and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store the data that has been output or will be output.

[0077] The above embodiments are only used to illustrate the design ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made according to the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. A method for predicting photoacoustic imaging array signals for limited field of view sampling, characterized in that, the method includes: Detect the acoustic pressure signal reflected by the tissue of the object to be measured through the ultrasonic transducer in the photoacoustic imaging array, convert the acoustic pressure signal into an electrical signal, and perform a stationarity test on the electrical signal; Select the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combine the truncation number and the trailing number in pairs to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion; Determine the result of the autoregressive moving average model according to the order, and then weight it with the result of the LSTM network model to construct a photoacoustic imaging array signal prediction model; wherein, the expression of the photoacoustic imaging array signal prediction model is: s k s(t) = αm k s(t) + βn k s(t) = α(γ 0 + γ 1 S t-1 + … + γ p S t-p + ε t + c 1 ε t-1 + … + c q ε t-q ) + βn k (t) Among them, s k (t) is the prediction result of the photoacoustic imaging array signal prediction model for the signal collected by the k-th ultrasonic transducer, m k (t) is the prediction result of the autoregressive moving average model for the signal collected by the k-th ultrasonic transducer, n k (t) is the prediction result of the LSTM network for the signal collected by the k-th ultrasonic transducer, t is the time t, α and β are weight coefficients, and α + β = 1; γ i is the autoregressive coefficient, i = 0, 1…p, S t-p is the input signal S t at the amplitude at time t - p, ε t-j is the time series of the error term, c j is the moving average coefficient, j = 1…q; Input the electrical signal obtained after the stationarity test into the photoacoustic imaging array signal prediction model for prediction, and end the prediction when the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, and obtain the predicted photoacoustic signals.

2. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 1, characterized in that, Performing a stationarity test on the electrical signal includes: The electrical signal is represented as an M×N matrix, where N is the length of the signal collected by a single ultrasonic transducer and M is the number of ultrasonic transducers; the matrix is unfolded, and the vector of size (1, M×N) is denoted as signal S t ; Perform ADF test and KPSS test on signal S t If the ADF value of the test result is equal to 1 and the KPSS value is equal to 0, the test is passed; if the ADF value of the test result is not equal to 0 or the KPSS value is not equal to 1, perform differential processing on signal S t until the stationarity test is passed.

3. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 1, characterized in that, Selecting the truncation number and the trailing number through the autocorrelation function and the partial autocorrelation function, and then combining the truncation number and the trailing number in pairs to determine the order of the photoacoustic imaging array signal prediction model based on the Bayesian information criterion includes: Select the truncation number and the trailing number by plotting the autocorrelation lag coefficient diagram and the partial correlation lag coefficient diagram, combine the truncation number and the trailing number in pairs to calculate the Bayesian information criterion of each combination, and obtain the Bayesian information criterion matrix; find the position of the smallest element in the BIC matrix, and record the position of this smallest element as the a-th row and the b-th column, and the order p of the photoacoustic imaging array signal prediction model = a - 1, q = b - 1.

4. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 1, characterized in that, the method further includes: Training the LSTM network, wherein, the optimizer is the root mean square backpropagation algorithm optimizer, the activation function is the Sigmoid function, and the loss function is the mean absolute error function; the LSTM network is composed of three layers of long short-term memory networks, a pooling layer and a fully connected layer.

5. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 1, characterized in that, Inputting the electrical signal obtained after the stationarity test into the photoacoustic imaging array signal prediction model for prediction includes: For the electrical signal S obtained after the stationarity test t perform L2 norm normalization to obtain signal S t * ; Input signal S t * into the photoacoustic imaging array signal prediction model, where the predicted value of the autoregressive moving average model is m k+1 (t), and the predicted value of the LSTM network is n k+1 (t). Then, the predicted value of the signal received by the ultrasonic transducer numbered k + 1 by the photoacoustic imaging array signal prediction model is: s k+1 (t) = αm k+1 (t) + βn k+1 (t); wherein, α and β are weight coefficients, and α + β = 1.

6. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 5, characterized in that, When the number of photoacoustic signals output by the photoacoustic imaging array signal prediction model reaches the preset number of ultrasonic transducers, ending the prediction and obtaining the predicted photoacoustic signals includes: Concatenate the predicted signal s k+1 (t) to the tail of the signal S t * to obtain the first concatenated signal, and input the first concatenated signal into the photoacoustic imaging array signal prediction model for prediction to obtain the predicted value s k+2 (t) of the signal received by the ultrasonic transducer numbered k + 2; Preset the number N of ultrasonic transducers in the photoacoustic imaging array, and end the prediction when the photoacoustic signal output by the photoacoustic imaging array signal prediction model is s N (t), and obtain the signal prediction value: S t * =(s 1 (t), s 2 (t), …, s k (t), s k+1 (t), …, s N (t)).

7. The method for predicting photoacoustic imaging array signals for limited field of view sampling according to claim 1 or 6, characterized in that, The signal length predicted by the photoacoustic imaging array signal prediction model each time is the same as the signal length collected by a single transducer in signal S. t ​ 8. An optoacoustic imaging array signal prediction device for limited field of view sampling, Characterized in that, It includes one or more processors for implementing the optoacoustic imaging array signal prediction method for limited field of view sampling according to any one of claims 1-7.

9. A computer-readable storage medium, on which a program is stored, Characterized in that, When the program is executed by a processor, it is used to implement the optoacoustic imaging array signal prediction method for limited field of view sampling according to any one of claims 1-7.

Images