Stack reconstruction imaging phase recovery method and device based on object function implicit representation

Abstract

The application relates to a layered reconstruction imaging phase recovery method and device based on an object function implicit representation, wherein the method comprises the following steps: generating an implicit representation of an object function by using a preset deep neural network; randomly sampling each pixel volume between adjacent slices, and calculating a corresponding projection potential function value according to the coordinates of the sampling points; performing forward simulation of the projection potential function value for multi-slice layered imaging; defining a corresponding loss function according to the difference between the simulation data and experimental data; judging whether the loss function meets a preset convergence condition; if the loss function meets the preset convergence condition, sampling in a three-dimensional space position by using the implicit representation and performing phase recovery of a three-dimensional object function to obtain object function information meeting a preset optimal condition. The embodiment of the application utilizes the representation capability of a neural network and the function continuity of the neural network in a solution space, avoids the short board of a local optimization trap of a traditional method, and improves the accuracy of phase recovery.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical imaging, in particular to a stacked reconstruction imaging phase recovery method and device based on implicit representation of object function. BACKGROUND

[0002] In the stacked imaging technology of multiple slices, the object function is usually represented by various functions on a two-dimensional surface at a specific spatial position, such as a pixel matrix, a two-dimensional function, etc.

[0003] In the related art, the regularization factor can be used to constrain the solution space characteristics or the inter-layer correlation to obtain a phase information reconstruction result.

[0004] However, the related art is prone to local optimal solution problems; at the same time, since the spatial position of the projection potential function needs to be preset and cannot be changed, this hyperparameter has an unavoidable impact on the algorithm. SUMMARY

[0005] The present application provides a stacked reconstruction imaging phase recovery method and device based on implicit representation of object function to solve the problem of reducing the accuracy of stacked imaging phase recovery in the related art.

[0006] The first aspect of the present application provides a stacked reconstruction imaging phase recovery method based on implicit representation of object function, comprising the following steps: generating an implicit representation of an object function by using a preset deep neural network; randomly sampling each pixel volume between adjacent slices to obtain a sampling point of each pixel volume, and calculating a corresponding projection potential function value according to the coordinates of the sampling point; performing forward simulation of the projection potential function value for multi-slice stacked imaging to obtain a forward simulation result, and obtaining simulation data of a series of convergent beams according to the forward simulation result; defining a corresponding loss function according to the difference between the simulation data and experimental data, and making the loss function satisfy a preset condition according to neural network parameters; in the case that the loss function satisfies the preset condition, judging whether the loss function satisfies a preset convergence condition; if the loss function satisfies the preset convergence condition, sampling in a three-dimensional space position and phase recovery of a three-dimensional object function by using the implicit representation to obtain object function information satisfying a preset optimal condition.

[0007] Optionally, in one embodiment of the present application, after judging whether the loss function satisfies the preset convergence condition, it further comprises: if the loss function does not satisfy the preset convergence condition, randomly or regularly sampling the slice thickness of the object function until the loss function satisfies the preset convergence condition, and stopping sampling.

[0008] Optionally, in an embodiment of the present application, the generating the implicit representation of the object function by using the deep neural network comprises: inputting the spatial coordinates or spatial coordinate encoding of the object function into the deep neural network to obtain the object function value corresponding to the spatial coordinate position according to the deep neural network.

[0009] Optionally, in an embodiment of the present application, the spatial coordinate encoding comprises at least one dimension expansion encoding in trigonometric function encoding and hash encoding.

[0010] Optionally, in an embodiment of the present application, the calculating the projection potential function value corresponding to the coordinate of the sampling point comprises: calculating the volume fraction of the sampling point according to the coordinate of the sampling point; multiplying the volume fraction and the potential function value of the sampling point to obtain a product result of the sampling point; accumulating the product result to obtain the projection potential of the next layer of pixels, and iteratively applying the projection potential to the preset slice layer to obtain the projection potential function value of the preset slice layer.

[0011] An embodiment of the second aspect of the present application provides a device for reconstructing imaging phase recovery based on implicit representation of an object function, comprising: a generating module configured to generate an implicit representation of an object function by using a preset deep neural network; a calculating module configured to randomly sample each pixel volume between adjacent slice layers to obtain a sampling point of the each pixel volume, and calculate a projection potential function value corresponding to the coordinate of the sampling point; a simulating module configured to perform forward simulation of multi-slice layer imaging on the projection potential function value to obtain a forward simulation result, and obtain simulation data of a series of convergent beams according to the forward simulation result; a defining module configured to define a corresponding loss function according to the difference between the simulation data and experimental data, and make the loss function satisfy a preset condition according to a neural network parameter; a judging module configured to judge whether the loss function satisfies a preset convergence condition in the case that the loss function satisfies the preset condition; and a phase recovery module configured to perform sampling in a three-dimensional space position and phase recovery of a three-dimensional object function by using the implicit representation when the loss function satisfies the preset convergence condition, to obtain object function information satisfying a preset optimal condition.

[0012] Optionally, in an embodiment of the present application, the device further comprises a sampling module configured to, after judging whether the loss function satisfies the preset convergence condition, randomly or regularly sample the slice thickness of the object function until the loss function satisfies the preset convergence condition in the case that the loss function does not satisfy the preset convergence condition, and stop sampling.

[0013] Optionally, in an embodiment of the present application, the generating module comprises a generating unit configured to input the spatial coordinates or the spatial coordinate encoding of the object function into the deep neural network to obtain the object function value corresponding to the spatial coordinate position according to the deep neural network.

[0014] Optionally, in an embodiment of the present application, the spatial coordinate encoding comprises at least one dimension expansion encoding in trigonometric function encoding and hash encoding.

[0015] Optionally, in an embodiment of the present application, the calculating module comprises a first calculating unit configured to calculate the volume fraction of the sampling point according to the coordinate of the sampling point, a second calculating unit configured to multiply the volume fraction and the potential function value of the sampling point to obtain a product result of the sampling point, and an accumulating unit configured to accumulate the product results to obtain the projection potential of the next layer of pixels, and iteratively apply the projection potential to the preset slice layer to obtain the projection potential function value of the preset slice layer.

[0016] A third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the object function implicit representation based stacked reconstruction imaging phase retrieval method according to the above embodiments.

[0017] A fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the object function implicit representation based stacked reconstruction imaging phase retrieval method according to the above embodiments.

[0018] A fifth aspect of the present application provides a computer program product, which stores a computer program, and the program is executed by a processor to implement the object function implicit representation based stacked reconstruction imaging phase retrieval method according to the above embodiments.

[0019] The embodiments of the present application utilize the representation capability of the neural network and the function continuity in the solution space, avoid the short board of the local optimization trap of the traditional method, and improve the accuracy of the phase retrieval. Thus, the problem of insufficient accuracy and efficiency of the stacked imaging phase retrieval in the related art is solved.

[0020] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings.

[0022] Figure 1 A flowchart of a layered reconstruction imaging phase retrieval method based on an implicit representation of an object function according to an embodiment of the present application is provided.

[0023] Figure 2 A full connection network input and output structure diagram of a layered reconstruction imaging phase retrieval method based on an implicit representation of an object function according to an embodiment of the present application is provided.

[0024] Figure 3 An algorithm flow diagram of a layered reconstruction imaging phase retrieval method based on an implicit representation of an object function according to an embodiment of the present application is provided.

[0025] Figure 4 A structure diagram of a layered reconstruction imaging phase retrieval device based on an implicit representation of an object function according to an embodiment of the present application is provided.

[0026] Figure 5 A structure diagram of an electronic device according to an embodiment of the present application is provided. DETAILED DESCRIPTION

[0027] Embodiments of the present application are described in detail below with reference to examples thereof shown in the attached drawings, wherein the same or similar numerals represent the same or similar elements throughout. The embodiments described below by reference to the drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0028] The layered reconstruction imaging phase retrieval method and device based on an implicit representation of an object function according to an embodiment of the present application are described below with reference to the drawings. In order to solve the problem of insufficient accuracy and efficiency of layered imaging phase retrieval in the related art mentioned above, the present application provides a layered reconstruction imaging phase retrieval method based on an implicit representation of an object function, in which the representation capability of a neural network and its function continuity in the solution space are utilized to avoid the short board of the local optimization trap of the traditional method, thereby improving the accuracy of phase retrieval. Thus, the problem of insufficient accuracy and efficiency of layered imaging phase retrieval in the related art is solved.

[0029] Specifically, Figure 1 A flow diagram of a layered reconstruction imaging phase retrieval method based on an implicit representation of an object function according to an embodiment of the present application is provided.

[0030] As Figure 1 shown, the layered reconstruction imaging phase retrieval method based on an implicit representation of an object function includes the following steps:

[0031] In step S101, a preset deep neural network is used to generate an implicit representation of an object function.

[0032] In actual implementation, the embodiments of this application can utilize a preset deep neural network to generate an implicit representation of the object function, that is, to construct a neural network model for implicitly representing the object function in multiple layers.

[0033] The embodiments of this application avoid the limitations of traditional regularization factors by using the implicit representation of neural networks, ensuring the global optimal solution in the multi-layer stacking reconstruction process, effectively overcoming the local optimization problem, improving the quality of imaging results, and improving the accuracy of phase recovery.

[0034] It should be noted that neural network models include, but are not limited to, fully connected neural networks, and should be selected and optimized according to specific imaging requirements.

[0035] Optionally, in one embodiment of this application, generating an implicit representation of a property function using a deep neural network includes: inputting the spatial coordinates or spatial coordinate encoding of the property function into the deep neural network to obtain the property function value corresponding to the spatial coordinate position based on the deep neural network.

[0036] As one possible implementation method, embodiments of this application can be constructed as follows: Figure 2 The fully connected neural network model F shown takes the spatial coordinates (X,Y,Z) of a point in the object function as input, the second layer of the neural network is a dimension expansion encoding of the three values, and the output of the neural network is the predicted value of the object function corresponding to the spatial coordinate position, which includes the real part A and the imaginary part B.

[0037] Optionally, in one embodiment of this application, the spatial coordinate encoding includes at least one dimension-extended encoding of trigonometric function encoding and hash encoding.

[0038] In this embodiment, the spatial coordinate encoding is a dimension-extended encoding, including but not limited to trigonometric function encoding and hash encoding.

[0039] In step S102, the volume of each pixel between adjacent layers is randomly sampled to obtain the sampling point of each pixel volume, and the corresponding projection potential function value is calculated based on the coordinates of the sampling point.

[0040] Understandably, the calculation of the projection potential function involves random or regular sampling and a weighted summation operation of its product with the volume fraction.

[0041] In actual implementation, the embodiments of this application can perform n random or regular samplings within the volume of the same pixel position between each pair of adjacent layers, and calculate based on the coordinates of the sampling points to finally obtain the corresponding projection potential function value.

[0042] The embodiments of this application utilize an innovative inter-pixel sampling mechanism to improve algorithm sampling efficiency, computational efficiency, and imaging quality.

[0043] Optionally, in one embodiment of this application, calculating the corresponding projection potential function value based on the coordinates of the sampling point includes: calculating the volume fraction of the sampling point based on the coordinates of the sampling point; multiplying the volume fraction and the potential function value of the sampling point to obtain the product result of the sampling point; accumulating the product result to obtain the projection potential of the next layer pixel; and iteratively applying the projection potential to a preset layer to obtain the projection potential function value of the preset layer.

[0044] It is understood that the pixel size and number in the embodiments of this application can be adjusted and optimized as needed during the iteration process; the preset layer in the embodiments of this application can be all layers; the embodiments of this application can perform multi-layer slicing on the three-dimensional material function, and the thickness of each layer in the embodiments of this application is randomly sampled to obtain the Z-direction coordinates {Z0, Z1, ..., Z...} of each layer. n}

[0045] Specifically, embodiments of this application can be applied to adjacent slices Z. i and Z i+1 Its interlaminar thickness is d i Set the pixel size and iterate through all pixels. Perform position sampling between layers for a given pixel [m, n], which means sampling the set {x, y, z | m}. <x<m+1,n<y<n+1,Z i <z<Z i+1 Sampling. In this embodiment, the sampling frequency is set to randomly divide the z-direction into 4 parts with a division length of {z1,z2,z3,z4}, and randomly sample {x,y,z} within each division region and calculate F(x,y,z) to obtain the potential function set {F1,F2,F3,F4}. The potential function values ​​at these sampling points are multiplied by the corresponding volume fraction to obtain the product result of the sampling points. The product results are accumulated to obtain the projection potential function value of the corresponding pixel in the next layer, i.e., P[m,n]=∑z i F i The above mechanism is applied iteratively to process all layers, obtaining the projection potential functions of all layers required for multi-layer simulation, and finally obtaining the projection potential function values.

[0046] The embodiments of this application utilize an innovative multi-layer thickness random sampling mechanism to further improve the algorithm's sampling efficiency, computational efficiency, and imaging quality.

[0047] In step S103, a forward simulation of multi-layer stacked imaging is performed on the projection potential function value to obtain the forward simulation results, and a series of convergent beam diffraction simulation data are obtained based on the forward simulation results.

[0048] It is understood that the interlayer spacing in the multi-layer simulation process in the embodiments of this application can be adjusted and optimized as needed during the iteration process.

[0049] In actual implementation, the embodiments of this application can refer to experimental conditions to complete the forward simulation process of multi-layer stacked imaging, and obtain a series of 4D simulation data of convergent beam diffraction, thereby providing support for improving the quality of subsequent imaging results.

[0050] In step S104, a corresponding loss function is defined based on the difference between the simulated data and the experimental data, and the loss function is made to meet the preset conditions based on the neural network parameters.

[0051] It is understood that in the embodiments of this application, satisfying the preset conditions for the loss function can be the design and minimization of the loss function.

[0052] In this embodiment, a loss function can be defined to measure the difference between simulated and experimental data. The defined loss function is the mean squared error (MSE). The neural network model parameters are optimized using a backpropagation algorithm to minimize this loss function and match the experimental data. This flexible and efficient imaging algorithm framework can be widely applied in various optical imaging fields, such as biomedicine and materials science.

[0053] The embodiments of this application can use the gradient descent algorithm to adjust the neural network parameters to reduce the loss, and utilize the backpropagation optimization strategy to further improve the algorithm's sampling efficiency, computational efficiency, and imaging quality.

[0054] It should be noted that the preset conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0055] In step S105, if the loss function meets the preset conditions, it is determined whether the loss function meets the preset convergence conditions.

[0056] In actual implementation, the embodiments of this application can determine whether the loss function has converged while minimizing the loss function, and then take targeted measures based on different convergence results.

[0057] It should be noted that the preset conditions and preset convergence conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0058] Optionally, in one embodiment of this application, after determining whether the loss function meets the preset convergence condition, the method further includes: if the loss function does not meet the preset convergence condition, randomly or regularly sampling the sheet thickness of the object function until the loss function meets the preset convergence condition and then stopping the sampling.

[0059] Specifically, in this embodiment of the application, when the loss function does not converge, the layer thickness of the object function can be randomly or regularly sampled, and the above steps of multi-layer slicing of the three-dimensional object function to define the loss function can be repeated until the loss function converges and sampling stops.

[0060] It should be noted that the preset convergence conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0061] In step S106, if the loss function satisfies the preset convergence condition, the implicit representation is used to sample at the three-dimensional spatial position and recover the phase of the three-dimensional object function to obtain the object function information that satisfies the preset optimal condition.

[0062] It is understood that the object function information that satisfies the preset optimal conditions in the embodiments of this application can be the object function information that satisfies high precision and global optimal solution.

[0063] In this embodiment of the application, when the loss function converges, it indicates that the implicit expression of the object function has reached the optimal state. At this time, the implicit expression can be used to sample the three-dimensional spatial position and recover the phase of the three-dimensional object function to obtain the object function information of high precision and global optimal solution.

[0064] The embodiments of this application can post-process the parameters of the neural network, such as full-space sampling, to obtain an explicit expression of the object function.

[0065] It should be noted that the preset optimal conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0066] Specifically, it can be combined with Figure 3 The working principle of the stacked reconstruction imaging phase retrieval method based on implicit representation of the object function in the embodiments of this application is illustrated in detail with a specific embodiment.

[0067] like Figure 3 As shown, embodiments of this application may include the following steps:

[0068] Step S301: Set up the neural network model.

[0069] Step S302: Sample the sheet thickness.

[0070] Step S303: Calculate the projection potential of each layer.

[0071] Step S304: Multi-layer forward simulation.

[0072] Step S305: Optimize the loss function.

[0073] Step S306: Determine whether the loss function has converged.

[0074] Step S307: End iteration, post-processing.

[0075] The phase retrieval method for stacked reconstruction imaging based on implicit representation of the object function proposed in this application can utilize the representational power of neural networks and their functional continuity in the solution space to avoid the shortcomings of local optimization traps in traditional methods, thereby improving the accuracy and efficiency of phase retrieval in stacked imaging. This solves the problem of insufficient accuracy and efficiency in phase retrieval of stacked imaging in related technologies.

[0076] Next, referring to the accompanying drawings, a phase recovery device for stacked reconstruction imaging based on implicit representation of the material function, according to an embodiment of this application, is described.

[0077] Figure 4 This is a schematic diagram of the structure of the stacked reconstruction imaging phase recovery device based on implicit representation of the object function according to an embodiment of this application.

[0078] like Figure 4 As shown, the stacked reconstruction imaging phase recovery device 10 based on the implicit representation of the object function includes: a generation module 100, a calculation module 200, a simulation module 300, a definition module 400, a judgment module 500, and a phase recovery module 600.

[0079] Specifically, the generation module 100 is used to generate implicit representations of object functions using a preset deep neural network.

[0080] The calculation module 200 is used to randomly sample the volume of each pixel between adjacent layers to obtain the sampling point of each pixel volume, and calculate the corresponding projection potential function value based on the coordinates of the sampling point.

[0081] The simulation module 300 is used to perform forward simulation of multi-layer stacked imaging of the projection potential function value to obtain forward simulation results, and to obtain a series of convergent beam diffraction simulation data based on the forward simulation results.

[0082] The definition module 400 is used to define the corresponding loss function based on the difference between the simulated data and the experimental data, and to make the loss function meet the preset conditions based on the neural network parameters.

[0083] The judgment module 500 is used to determine whether the loss function meets the preset convergence condition, provided that the loss function meets the preset condition.

[0084] The phase recovery module 600 is used to sample and recover the phase of the three-dimensional object function in three-dimensional space using implicit representation when the loss function meets the preset convergence conditions, so as to obtain the object function information that meets the preset optimal conditions.

[0085] Optionally, in one embodiment of this application, the stacked reconstruction imaging phase recovery device 10 based on the implicit representation of the object function further includes a sampling module.

[0086] The sampling module is used to randomly or regularly sample the sheet thickness of the object function after determining whether the loss function meets the preset convergence condition. If the loss function does not meet the preset convergence condition, the sampling stops.

[0087] Optionally, in one embodiment of this application, the generation module 100 includes a generation unit.

[0088] The generation unit is used to input the spatial coordinates or spatial coordinate encoding of the object function into the deep neural network, so as to obtain the object function value corresponding to the spatial coordinate position according to the deep neural network.

[0089] Optionally, in one embodiment of this application, the spatial coordinate encoding includes at least one dimension-extended encoding of trigonometric function encoding and hash encoding.

[0090] Optionally, in one embodiment of this application, the calculation module 200 includes: a first calculation unit, a second calculation unit, and an accumulation unit.

[0091] The first calculation unit is used to calculate the volume fraction of the sampling point based on the coordinates of the sampling point.

[0092] The second calculation unit is used to multiply the volume fraction and the potential function value of the sampling point to obtain the product result of the sampling point.

[0093] The accumulation unit is used to accumulate the product results to obtain the projection potential of the next layer pixel, and to iteratively apply the projection potential to the preset layer to obtain the projection potential function value of the preset layer.

[0094] It should be noted that the foregoing explanation of the embodiment of the stacked reconstruction imaging phase retrieval method based on implicit representation of the object function also applies to the stacked reconstruction imaging phase retrieval device based on implicit representation of the object function in this embodiment, and will not be repeated here.

[0095] The phase retrieval device for stacked reconstruction imaging based on implicit representation of the object function proposed in this application can utilize the representational power of neural networks and their functional continuity in the solution space to avoid the shortcomings of local optimization traps in traditional methods, thereby improving the accuracy of phase retrieval. This solves the problem of insufficient accuracy and efficiency in phase retrieval of stacked imaging in related technologies.

[0096] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0097] The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0098] When the processor 502 executes the program, it implements the stacked reconstruction imaging phase recovery method based on implicit representation of the object function provided in the above embodiments.

[0099] Furthermore, electronic devices also include:

[0100] Communication interface 503 is used for communication between memory 501 and processor 502.

[0101] The memory 501 is used to store computer programs that can run on the processor 502.

[0102] The memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0103] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0104] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0105] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0106] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described stacked reconstruction imaging phase retrieval method based on implicit representation of the object function.

[0107] This application also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described stacked reconstruction imaging phase retrieval method based on implicit representation of the object function.

[0108] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0109] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0110] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0112] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0113] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0114] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0115] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for phase retrieval in tomographic reconstruction based on an object function implicit representation, characterized in that, The method comprises the following steps: generating an implicit representation of an object function by using a preset deep neural network; randomly sampling each pixel volume between adjacent slices to obtain a sampling point of the each pixel volume, and calculating a corresponding projection potential function value according to the coordinates of the sampling point; performing forward simulation on the projection potential function value for multi-slice layer imaging to obtain a forward simulation result, and obtaining simulated data of series convergent beam diffraction according to the forward simulation result; defining a corresponding loss function according to the difference between the simulated data and experimental data, and making the loss function satisfy a preset condition according to a neural network parameter; judging whether the loss function satisfies a preset convergence condition in the case that the loss function satisfies the preset condition; if the loss function satisfies the preset convergence condition, sampling in a three-dimensional space position and phase recovery of a three-dimensional object function are performed by using the implicit representation to obtain object function information satisfying a preset optimal condition.

2. The method of claim 1, wherein, After judging whether the loss function satisfies the preset convergence condition, the method further comprises: if the loss function does not satisfy the preset convergence condition, randomly or regularly sampling a slice thickness of the object function until the loss function satisfies the preset convergence condition, and stopping sampling.

3. The method of claim 1, wherein, The generating of the implicit representation of the object function by using the preset deep neural network comprises: inputting spatial coordinates or spatial coordinate encoding of the object function into the deep neural network to obtain an object function value corresponding to a spatial coordinate position according to the deep neural network.

4. The method of claim 3, wherein, The spatial coordinate encoding comprises at least one dimension extension encoding in trigonometric function encoding and hash encoding.

5. The method of claim 1, wherein, The calculation of the corresponding projection potential function value according to the coordinates of the sampling point comprises: calculating a volume fraction of the sampling point according to the coordinates of the sampling point; multiplying the volume fraction and a potential function value of the sampling point to obtain a product result of the sampling point; accumulating the product result to obtain a projection potential of a next layer of pixels, and iteratively applying the projection potential to a preset slice to obtain a projection potential function value of the preset slice.

6. A stacked reconstruction imaging phase retrieval apparatus based on object function implicit representation, characterized in that, The method comprises: a generating module configured to generate an implicit representation of an object function by using a preset deep neural network; a calculating module configured to randomly sample each pixel volume between adjacent slices to obtain a sampling point of the each pixel volume, and calculate a corresponding projection potential function value according to the coordinates of the sampling point; a simulating module configured to perform forward simulation on the projection potential function value for multi-slice layer imaging to obtain a forward simulation result, and obtain simulated data of series convergent beam diffraction according to the forward simulation result; a defining module configured to define a corresponding loss function according to the difference between the simulated data and experimental data, and make the loss function satisfy a preset condition according to a neural network parameter; a judging module configured to judge whether the loss function satisfies a preset convergence condition in the case that the loss function satisfies the preset condition; and if the loss function satisfies the preset convergence condition, sampling in a three-dimensional space position and phase recovery of a three-dimensional object function are performed by using the implicit representation to obtain object function information satisfying a preset optimal condition. The phase recovery module is configured to, when the loss function meets the preset convergence condition, perform phase recovery of the three-dimensional space position sampling and the three-dimensional object function by using the implicit representation, to obtain object function information meeting a preset optimization condition.

7. The apparatus of claim 6, wherein, Further comprising: The sampling module is configured to, after judging whether the loss function meets the preset convergence condition, randomly or regularly sample a slice thickness of the object function until the loss function meets the preset convergence condition, and stop sampling.

8. The apparatus of claim 6, wherein, The generation module comprises: The generation unit is configured to input the spatial coordinates or spatial coordinate encoding of the object function into the deep neural network, to obtain an object function value corresponding to a spatial coordinate position according to the deep neural network.

9. The apparatus of claim 8, wherein, The spatial coordinate encoding comprises at least one dimension expansion encoding in a trigonometric function encoding and a hash encoding.

10. The apparatus of claim 6, wherein, The calculation module comprises: The first calculation unit is configured to calculate a volume fraction of the sampling point according to the coordinate of the sampling point; The second calculation unit is configured to multiply the volume fraction and a potential function value of the sampling point, to obtain a product result of the sampling point; The accumulation unit is configured to accumulate the product result, to obtain a projection potential of a next layer of pixels, and iteratively apply a preset slice according to the projection potential, to obtain a projection potential function value of the preset slice.

11. An electronic device, comprising: Comprise: A memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the method for reconstructing imaging phase recovery based on an implicit representation of an object function according to any one of claims 1-5.

12. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the method for reconstructing imaging phase recovery based on an implicit representation of an object function according to any one of claims 1-5.

13. A computer program product comprising a computer program, characterized in that, The computer program is executed to implement the method for reconstructing imaging phase recovery based on an implicit representation of an object function according to any one of claims 1-5.

Images