Deep learning based method and system for fast generation of transcranial focused acoustic holography lens

Abstract

The application discloses a kind of fast generation method and system of transcranial focused acoustic holographic lens based on deep learning, uses k-wave toolbox in Matlab, utilizes time reversal method to set the focus point as sound source, captures the phase at the plane of ultrasonic transducer, to design acoustic holographic lens;With the position of preset focus point and the holographic lens designed as data set training U-net network, the input of the U-net network is the position of preset focus point, and the output is the holographic lens designed;Set target focus point, input the U-net network trained, and obtain the holographic lens corresponding to the target focus point from the U-net network output trained;Based on the U-net network output holographic lens design, the holographic lens is manufactured by 3D printing.Based on the framework of deep learning, fast and accurate acoustic holographic lens is generated, and the problem that the acoustic holographic lens is consumed too large computing resources in the process of transcranial focusing by acoustic holographic lens is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical imaging, and particularly relates to a transcranial focused acoustic holographic lens fast generation method and system based on deep learning. BACKGROUND

[0002] Transcranial focused ultrasound has great development potential due to its non-invasive, safe, deep penetration and other advantages. The ultrasonic wave can maintain a high penetration depth when passing through the skull, so that it can directly affect the deep brain tissue, and the heat effect generated by the focused ultrasound helps to treat deep diseases. Therefore, compared with the traditional transcranial surgical treatment method, the transcranial focused ultrasound technology can realize surgical treatment without craniotomy, thereby reducing the risk of infection and bleeding, and has a wide application scenario in the biomedical field. For example, at low intensity, transcranial focused ultrasound can reversibly open the blood-brain barrier without tissue ablation, and at high intensity, transcranial focused ultrasound can produce tissue ablation. However, when the ultrasound passes through the skull, due to the nature and structural complexity of the skull, it will cause strong ultrasonic scattering and attenuation, resulting in serious distortion and distortion of the transcranial sound field. At present, the transcranial focusing by modulating the phase of the emitted plane wave through the ultrasonic holographic lens is a transcranial focusing method with stable effect and low cost. The holographic lens used for transcranial focusing is currently designed based on the time reversal method, which needs to perform three-dimensional simulation in the process of designing the hologram, and the calculation cost is large, thereby limiting the flexibility and practicality of the application of the acoustic holographic lens in transcranial focused ultrasound. SUMMARY

[0003] To solve the above problems, the purpose of the present application is to provide a transcranial focused acoustic holographic lens fast generation method and system based on deep learning, which realizes fast and accurate acoustic holographic lens generation based on a deep learning framework, solves the problem of excessive consumption of computing resources in the process of transcranial focusing by acoustic holographic lens, and can generate acoustic holographic lens in real time.

[0004] The technical scheme provided by the present application is: a transcranial focused acoustic holographic lens fast generation method based on deep learning, comprising the following steps:

[0005] Step S1: using the k-wave toolbox in Matlab, setting a preset focal point as a sound source by using the time reversal method, capturing the phase at the plane of the ultrasonic transducer, and designing an acoustic holographic lens;

[0006] Step S2: using the position of the preset focal point in step S1 and the designed holographic lens as a data set to train a U-net network, wherein the input of the U-net network is the position of the preset focal point, and the output is the designed holographic lens;

[0007] Step S3, setting a target focus point, inputting the trained U-net network, and outputting a holographic lens corresponding to the target focus point from the trained U-net network;

[0008] Step S4, based on the U-net network output holographic lens design, manufacturing the holographic lens by 3D printing.

[0009] Preferably, in step S1, the acoustic holographic lens is designed by time reversal method, the target focus point position is preset first, the focus point is taken as a sound source in simulation, signals are received at the transducer placement position through the skull, the phase of the signals is extracted, and the extracted phase information is converted into the height spectrum of each pixel of the acoustic holographic lens according to the following formula: h(i,j)=w(i,j).c water. c lens / 2πf.(c lens -c water ), wherein w(i,j) is the phase corresponding to the pixel (i,j), c water and c lens are the sound speeds in water and the holographic lens material, and f is the center frequency of the ultrasound, so as to obtain the height of each pixel of the holographic lens.

[0010] Preferably, the U-net network is composed of an up-sampling module and a down-sampling module, the down-sampling module reduces the dimension of a fixed-size image to make it conform to the size of a display area, and extracts deeper image features, the up-sampling module enlarges the image based on this, and fuses features of each layer by copying and cropping, and the U-net network jump connection can effectively fuse deep and shallow layer information of each scale, thereby improving the network training efficiency.

[0011] Preferably, the loss function of the U-net network is: wherein is the nth hologram calculated by the time reversal method, as a true value, is a hologram output by the corresponding U-net network, N represents the number of target images in a small batch training scale, an optimizer is used to minimize the loss function in the training process, and the network parameters are automatically updated after iteration optimization after each training.

[0012] Preferably, the U-net network is trained, the target focus point image is taken as the input of the U-net network, and the Ground Truth is the hologram generated by the time reversal method as the output of the U-net network, in simulation, the image size is 64x64 pixels, and the pixel size of a single pixel is 750μm, and the size is 48mmx48mm.

[0013] Preferably, the implementation of the U-net network framework uses Tensorflow2.13 and Python 3.8. The U-net network framework is trained and tested on a Tesla V100-16G graphics processor, the encoder is optimized using the Adam optimizer, the learning rate is set to 0.001, the mini-batch is set to 16, and the training is performed 1500 times.

[0014] Preferably, after the U-net network training is completed, the U-net network is verified with the data in the test set, and the structural similarity SSIM between the acoustic hologram output by the U-net network and the acoustic hologram generated by the time reversal method is calculated. The formula for calculating the structural similarity SSIM is:

[0015] Wherein μ x is the acoustic hologram generated by the proposed network, μ y is the acoustic hologram generated by the time reversal method, σ x and σ y respectively represent the variance of each other, σ xy then indicates the covariance between the two images, c1 and c2 are two regularization parameters, the introduction of the two parameters is to ensure that the structural similarity SSIM does not error when encountering 0 value during calculation, and the numerical stability of the calculation is improved, c1 and c2 are defined as c1=(α×L) 2 , c2=(β×L) 2 , wherein L=64, α=0.01, β=0.03.

[0016] Based on the same idea, the application also provides a deep learning-based transcranial focused acoustic holographic lens fast generation system, comprising:

[0017] A data acquisition module is used to use the k-wave toolbox in Matlab, set the preset focus point as a sound source by using the time reversal method, capture the phase at the ultrasonic transducer plane, and design an acoustic holographic lens;

[0018] A training module is used to train a U-net network with the position of the preset focus point in the data acquisition module and the designed holographic lens as a data set, wherein the input of the U-net network is the position of the preset focus point, and the output is the designed holographic lens.

[0019] A lens generation module is used to set a target focus point, input the trained U-net network, and obtain the holographic lens corresponding to the target focus point from the output of the trained U-net network.

[0020] A lens printing module is used to manufacture the holographic lens by 3D printing based on the U-net network output holographic lens design.

[0021] The application also provides an electronic device based on the same concept, comprising:

[0022] A memory is used to store a processing program;

[0023] A processor is used to implement the deep learning-based fast generation method of the transcranial focused acoustic holographic lens according to any one of the above embodiments when the processing program is executed.

[0024] The application also provides a readable storage medium based on the same concept, wherein the readable storage medium stores a processing program, and the processing program is executed by a processor to implement the deep learning-based fast generation method of the transcranial focused acoustic holographic lens according to any one of the above embodiments.

[0025] Compared with the prior art, the application has the following advantages and positive effects:

[0026] The network parameters are trained by the deep learning method to replace the three-dimensional simulation in the computer to improve the calculation speed. It has been verified that the effect of reconstructing the focal point is equivalent to that of the traditional time reversal method, and the acoustic holographic lens for transcranial focusing can be quickly manufactured. In the embodiment, the position of the preset focal point is used to calculate the data of the acoustic holographic lens for transcranial focusing by the deep learning model at a speed of tens of milliseconds, and the reconstruction result of the acoustic holographic lens generated by the method is equivalent to that of the acoustic holographic lens generated by the traditional time reversal method, and the generation speed is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0027] The specific embodiments of the application will be further described in detail below with reference to the accompanying drawings, in which:

[0028] Figure 1 The embodiment of the application is a deep learning-based fast generation method of a transcranial focused acoustic holographic lens, and a flowchart thereof is shown in the figure.

[0029] Figure 2 The U-net network structure of the application is shown in the figure.

[0030] Figure 3 The holographic surface phase information used to manufacture the acoustic holographic lens is shown in the figure.

[0031] Figure 4 The focusing effect of the holographic lens generated in real time by the U-net network in simulation and experiment is shown in the figure. DETAILED DESCRIPTION

[0032] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer from the following description and claims. It should be noted that the drawings are all in a very simplified form and use non-precise ratios, and are only used to facilitate and clarify the illustration of the embodiments of the present invention.

[0033] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0034] First Embodiment

[0035] like Figure 1 As shown in this embodiment, a method for rapid generation of transcranial focused acoustic holographic lenses based on deep learning includes the following steps:

[0036] Step S1: In Matlab, use the k-wave toolbox and the time reversal method to set the preset focus point as the sound source, capture the phase at the plane of the ultrasonic transducer, and design the acoustic holographic lens accordingly.

[0037] Step S2: Use the preset focal position and the designed holographic lens from step S1 as the dataset to train the U-net network. The input of the U-net network is the preset focal position, and the output is the designed holographic lens.

[0038] Step S3: Set the target focal point, input the trained U-net network, and obtain the holographic lens corresponding to the target focal point from the output of the trained U-net network;

[0039] Step S4: Based on the holographic lens design output by the U-net network, the holographic lens is manufactured using 3D printing.

[0040] This invention trains network parameters using deep learning to replace 3D simulation in a computer, thus improving computational speed. Verification has shown that the effect of reconstructing the focal point is comparable to that of the traditional time-reversal method, and it can rapidly fabricate acoustic holographic lenses for transcranial focusing. In this embodiment, the data for the acoustic holographic lens used in transcranial focusing can be generated at a computational speed of tens of milliseconds using a deep learning model based on a preset focal point position. Comparison shows that the reconstruction results of the acoustic holographic lens generated by this method are comparable to those generated by the traditional time-reversal method, while significantly improving the generation speed.

[0041] Preferably, in step S1, the acoustic holographic lens is designed by time reversal method, the target focusing focal point position is preset, the focal point is taken as a sound source in simulation, signals are received at the position where the ultrasonic transducer is placed through the skull, the phase of the signals is extracted, and the extracted phase information is converted into the height spectrum of each pixel of the acoustic holographic lens according to the following formula: h(i,j) = w(i,j) c water .c lens / 2πf.(c lens -c water ), wherein w(i,j) is the phase corresponding to the pixel (i,j), c water and c lens are the sound speeds in water and the holographic lens material, and f is the central frequency of the ultrasound, so as to obtain the height of each pixel of the holographic lens.

[0042] In step S1, the k-wave toolbox in Matlab is used to design the acoustic holographic lens by time reversal method, the preset focusing focal point is taken as a sound source, and the phase at the position where the ultrasonic transducer is placed is captured. In the simulation process of the embodiment, the grid pixel size is set to 250 um, the simulation area size is 192*192*220 pixels, the ultrasonic central frequency is 2 MHz, and the Courant number is 0.2. Based on this, the holographic lens is designed by time reversal method. First, the target focusing focal point position is preset, the focal point is taken as a sound source in simulation, signals are received at the position where the ultrasonic transducer is placed after the skull, the phase of the signals is extracted, and the phase of each pixel point on the holographic surface is set to the arithmetic average of the surrounding three pixel points. The position of the preset focal point in step S1 and the designed holographic lens are taken as a data set, the U-net network is trained, the input is the position of the preset focal point, and the output is a phase map. The height of each pixel point of the acoustic holographic lens can be calculated through the phase map.

[0043] Preferably, the U-net network is composed of an up-sampling module and a down-sampling module. The down-sampling module reduces the dimension of a fixed-size image to make it conform to the size of a display area, and takes deeper image features. The up-sampling module enlarges the image based on this, and fuses features of each layer by copying and cropping. The U-net network jump connection can effectively fuse deep and shallow layer information of each scale, thereby improving the network training efficiency.

[0044] The U-net network architecture is as follows: Figure 2As shown, it is mainly composed of upsampling and downsampling, the main idea is to reduce the dimension of the fixed size image to make it conform to the size of the display area, and generate a thumbnail of the corresponding image to extract deeper image features. Then, use upsampling to enlarge the image, and fuse the features of each layer of downsampling and upsampling by copying and cropping. Finally, the convolutional layer can learn to combine more accurate outputs from these information. An important advantage of the U-net network is that the upsampling part has a large number of feature channels, which enables the network to propagate contextual information to higher resolution layers. We prepared 10000 sets of data sets, of which 9000 sets were used as the training set and 1000 sets were used as the test set, the size of the input and output was 64x64 pixels, the network was trained, and the loss function was MSE loss function.

[0045] Preferably, the loss function of the U-net network is: wherein is the nth hologram calculated by the time reversal method, as the true value, is the hologram output by the corresponding U-net network, N represents the target image number in the small batch training scale, the optimizer is used to minimize the loss function during the training process, and the network parameters are automatically updated after each training.

[0046] Mean Squared Error (MSE) is used as the loss function in the network, the Mean Squared Error loss function minimizes the squared difference between the predicted value and the true value to guide the training process of the network, so that the similarity between the generated hologram and the target image is maximized. Through this method, high-quality holograms suitable for transcranial focusing can be effectively generated. In the training process, the optimizer is used to minimize the loss function. After each training, the network parameters are automatically updated after iterative optimization.

[0047] Preferably, the U-net network is trained, the target focal image is used as the input of the U-net network, and the Ground Truth is the hologram generated by the time reversal method as the output of the U-net network. In the simulation, the image size is 64x64 pixels, and the pixel size of a single pixel is 750μm, and the size is 48mmx48mm.

[0048] Preferably, the implementation of the U-net network framework uses Tensorflow2.13 and Python 3.8. The U-net network framework is trained and tested on a Tesla V100-16G graphics processor, the encoder is optimized using Adam optimizer, the learning rate is set to 0.001, the mini batch is set to 16, and the training is performed 1500 times.

[0049] To train the network, 20000 pairs of data were prepared. Each set of data includes a target focal image and an acoustic hologram generated at a frequency of 2MHz using the time reversal method (as Ground Truth). The network takes the target focal image as input and outputs the corresponding acoustic hologram, which is applicable to the ultrasonic frequency of 2MHz. For this network, the target focal image (a binary image of 64x64) is the input of the network, and the Ground Truth is the hologram generated by the time reversal method. Among them, 18000 sets of data are used as the training set to train the network, and 2000 sets of data are used as the test set. In the simulation, the image size is 64x64 pixels, and the pixel size of a single pixel is 750pm, so the actual size is 48mmx48mm. The implementation of the network framework uses Tensorflow2.13

[78] and Python 3.8. The framework is trained and tested on a Tesla V100-16G graphics processing unit (GPU). The encoder is optimized using the Adam optimizer

[79] , with a learning rate of 0.001 and a mini-batch of 16. Training is performed for 1500 times.

[0050] Preferably, after the training of the U-net network is completed, the data in the test set are used to verify the U-net network, and the structural similarity SSIM between the acoustic hologram output by the U-net network and the acoustic hologram generated by the time reversal method is calculated. The formula for calculating the structural similarity SSIM is:

[0051] Wherein μ x is the acoustic hologram generated by the proposed network, μ y is the acoustic hologram generated by the time reversal method, σ x and σ y represent the variances of each other, σ xy represents the covariance between the two images, and c1 and c2 are two regularization parameters. These two parameters are introduced to ensure that the structural similarity SSIM does not encounter zero values during calculation, improving the numerical stability of the calculation. c1 and c2 are defined as c1=(α×L) 2 , c2=(β×L) 2 , where L=64, α=0.01, β=0.03.

[0052] With the increase of the number of cycle iterations, the loss function gradually converges, and after 1500 cycle iterations, the model has completely converged. The SSIM value ranges from -1 to 1, and the closer to 1 represents the higher similarity of the two images. In the performance of 2000 pairs of test sets, the average SSIM value is 0.972.

[0053] Step S3, select several groups of holographic lenses generated by the preset focus point through the network for verification in the simulation, and the phase diagram of the generated acoustic holographic lens is as shown in Figure 3 The phase diagram is then converted into the height of each pixel of the acoustic holographic lens, and is manufactured through 3D printing.

[0054] In step S4, the settings in the simulation are consistent with step S1, and in the experiment, the size of the ultrasonic transducer is 5cm*5cm, and a sinusoidal wave is emitted at 2MHz. The phase of the sinusoidal wave is modulated after passing through the holographic lens, and then focused behind the skull after passing through the skull. The simulation and experimental results after reconstruction are as shown in Figure 4 It can be seen that the positions in the numerical simulation and the real experiment are basically the same. The SSIM value and the PSNR value of the normalized acoustic field intensity of the target point position and the simulation and the experimental reconstruction are selected as the quantitative indicators to evaluate the quality of the reconstructed acoustic field. In the simulation, the SSIM value of the reconstruction result of the acoustic holographic lens made by the time reversal method based on the traditional method is 0.91, and the PSNR value is 32.69. The SSIM value of the reconstruction result of the acoustic holographic lens made by the U-net method is 0.82, and the PSNR value is 32.94. In the experiment, the SSIM value of the reconstruction result of the acoustic holographic lens made by the traditional time reversal method is 0.81, and the PSNR value is 27.47. The reconstruction result of the acoustic holographic lens made by the U-net deep learning method is 0.74, and the PSNR value is 26.96. It is not difficult to see that the performances of the two in the reconstruction of the transcranial focused acoustic field are quite comparable. However, it takes about 120s to generate a piece of acoustic holographic lens for transcranial focusing based on the traditional time reversal method in the computing environment of NVIDIA Tesla V100-16G, but it only takes 15ms to generate a piece of acoustic holographic lens for transcranial focusing based on the U-net method, and the generation speed is improved by several orders of magnitude.

[0055] Second embodiment

[0056] Based on the same concept, the application also provides a deep learning-based fast generation system of transcranial focusing acoustic holographic lens, comprising:

[0057] The data acquisition module is configured to set a preset focal point as a sound source by using a time reversal method in Matlab by using a k-wave toolbox, and capture a phase at a plane of an ultrasonic transducer to design an acoustic holographic lens.

[0058] The training module is configured to train a U-net network by taking a position of the preset focal point in the data acquisition module and the designed holographic lens as a data set, wherein an input of the U-net network is the position of the preset focal point, and an output of the U-net network is the designed holographic lens.

[0059] The lens generation module is configured to set a target focal point, input the trained U-net network, and output a holographic lens corresponding to the target focal point from the trained U-net network.

[0060] The lens printing module is configured to manufacture the holographic lens by 3D printing based on the holographic lens design output by the U-net network.

[0061] Based on the same concept, the application further provides an electronic device, which comprises a memory configured to store a processing program and a processor configured to implement the deep learning-based transcranial focused acoustic holographic lens fast generation method according to any one of the above embodiments when executing the processing program.

[0062] Based on the same concept, the application further provides a readable storage medium, which stores a processing program, and the processing program is executed by a processor to implement the deep learning-based transcranial focused acoustic holographic lens fast generation method according to any one of the above embodiments.

[0063] The deep learning-based transcranial focused acoustic holographic lens fast generation method can be stored in a computer-readable storage medium if it is implemented in the form of program instructions and sold or used as an independent product. Based on this understanding, the technical solution of the embodiment or the whole or part of the technical solution that essentially contributes to the prior art can be embodied in the form of software, which is stored in a storage medium and includes a plurality of instructions for making a computer device (which can be a personal computer, a server, or a network device) execute all or part of the steps of the method described in the embodiments of the disclosure. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0064] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific execution of the system and the device described above can refer to the corresponding process in the foregoing method embodiment.

[0065] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments. Even if various changes are made to the present application, if the changes belong to the scope of the claims of the present application and the equivalent technology thereof, they still fall within the protection scope of the present application.

Claims

1. A method for fast generation of a transcranial focused acoustic holography lens based on deep learning, characterized in that, The method comprises the following steps: Step S1, in Matlab, using the k-wave toolbox, setting a preset focus point as a sound source by using the time reversal method, capturing the phase at the plane of the ultrasonic transducer, and designing an acoustic holographic lens; Step S2, using the position of the preset focus point in step S1 and the designed holographic lens as a data set to train a U-net network, wherein the input of the U-net network is the position of the preset focus point, and the output is the designed holographic lens; Step S3, setting a target focus point, inputting the trained U-net network, and obtaining the holographic lens corresponding to the target focus point from the output of the trained U-net network; Step S4, based on the U-net network output holographic lens design, manufacturing the holographic lens through 3D printing; In S1, the acoustic holographic lens is designed by time reversal method, the target focus point position is preset, the focus point is taken as a sound source in simulation, signals are received at the transducer placement position through the skull, the phase of the signals is extracted, and the extracted phase information is converted into the height spectrum of each pixel of the acoustic holographic lens according to the following formula: h(i,j) = w(i,j).c water .c lens / 2πf.(c lens -c water ), wherein w(i,j) is the phase corresponding to the pixel (i,j), c water and c lens are the sound speeds in water and the holographic lens material, and f is the central frequency of the ultrasound, so as to obtain the height of each pixel of the holographic lens.

2. The method of claim 1, wherein the method is a deep learning based fast generation method of a transcranial focused acoustic holography lens. The U-net network is composed of an up-sampling module and a down-sampling module, the down-sampling module reduces the dimension of a fixed-size image to make it conform to the size of a display area, and extracts deeper image features, the up-sampling module enlarges the image based on the features, and fuses features of different layers through copying and cropping, and the U-net network jump connection can effectively fuse deep and shallow layer information of different scales, thereby improving the network training efficiency.

3. The method of claim 2, wherein the method is a deep learning based fast generation method of a transcranial focused acoustic holography lens. The loss function of the U-net network is: wherein is the nth hologram calculated by the time reversal method, as a true value, is the hologram output by the corresponding U-net network, N represents the target image number in the small batch training scale, and an optimizer is used in the training process to minimize the loss function. After each training, the network parameters are automatically updated after iterative optimization.

4. The method of claim 3, wherein the method is a deep learning-based fast generation method of a transcranial focused acoustic holography lens. The U-net network is trained, the target focus point image is used as the input of the U-net network, and the GroundTruth is the holographic lens generated by the time reversal method, in simulation, the image size is 64*64 pixels, and the pixel size of a single pixel is 750 mu m, and the size is 48 mm*48 mm.

5. The method of claim 4, wherein the method is a deep learning based fast generation method of a transcranial focused acoustic holography lens. The U-net network framework is realized by using Tensorflow2.13 and Python3.8, the U-net network framework is trained and tested on a TeslaV100-16G graphics processor, the encoder is optimized by using the Adam optimizer, the learning rate is set to 0.001, the mini batch is set to 16, and the training is performed 1500 times. 6.The method of claim 1, wherein, After the training of the U-net network is completed, the data in the test set are used to verify the U-net network, the structural similarity SSIM between the acoustic holographic lens output by the U-net network and the acoustic holographic lens generated by the time reversal method is calculated, and the calculation formula of the structural similarity SSIM is: where μ x is the sound hologram generated by the proposed network, μ y is the sound hologram generated by the time reversal method, σ x and σ y represent their respective variances, σ xy denotes the covariance between the two images, and c1 and c2 are two regularization parameters, which are introduced to ensure that the structural similarity SSIM does not encounter a value of 0 when calculating, thereby improving the numerical stability of the calculation, and c1 and c2 are defined as c1 = (α × L) 2 , c2 = (β × L) 2 , where L = 64, α = 0.01, and β = 0.

03.

7. A deep learning-based fast generation system of a transcranial focused acoustic holographic lens, characterized in that, The method comprises the following steps: A data acquisition module is configured to use the k-wave toolbox in Matlab, set a preset focus point as a sound source by using the time reversal method, capture the phase at the plane of the ultrasonic transducer, and design an acoustic holographic lens; A training module is configured to use the position of the preset focus point in the data acquisition module and the designed holographic lens as a data set to train a U-net network, wherein the input of the U-net network is the position of the preset focus point, and the output is the designed holographic lens; A lens generation module is configured to set a target focus point, input the trained U-net network, and obtain the holographic lens corresponding to the target focus point from the output of the trained U-net network. A lens printing module is used to manufacture the holographic lens based on the U-net network output holographic lens design through 3D printing. Wherein, the data acquisition module designs the acoustic holographic lens through time reversal method, first presets the target focusing focal point position, in the simulation, the focal point is taken as the sound source, the signal is received at the ultrasonic transducer placement position through the skull, the phase is extracted, the phase information extracted is converted into the height spectrum of each pixel of the acoustic holographic lens according to the following formula, h(i,j) = w(i,j).c water .c lens / 2πf.(c lens -c water ), wherein w(i,j) is the phase corresponding to the pixel (i,j), c water and c lens are the sound speeds in water and the holographic lens material, and f is the center frequency of the ultrasonic, so as to obtain the height of each pixel of the holographic lens.

8. An electronic device, comprising: The application relates to a method for quickly generating a transcranial focused acoustic holographic lens based on deep learning, and a device thereof. The memory is used for storing a processing program. The processor realizes the method for quickly generating a transcranial focused acoustic holographic lens based on deep learning as claimed in any one of claims 1 to 6 when executing the processing program.

9. A readable storage medium, characterized by, The readable storage medium stores a processing program, and the processing program realizes the method for quickly generating a transcranial focused acoustic holographic lens based on deep learning as claimed in any one of claims 1 to 6 when executed by a processor.

Images