Diffraction imaging method based on phase retrieval

Through the diffraction imaging method based on phase recovery, deep learning is used to process the phase change information in the X-ray diffraction pattern, the problem of poor resolution of traditional X-ray imaging is solved, and the imaging effect with higher contrast and resolution is achieved.

CN118566275BActive Publication Date: 2024-11-19NANTONG XIFENG MEDICAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411027739.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-19
Estimated Expiration
2044-07-30

AI Technical Summary

Technical Problem

Traditional X-ray imaging technology mainly relies on the intensity changes after X-ray penetrates an object, resulting in poor imaging resolution and it is difficult to provide sufficient internal structural information.

Method used

Using a diffraction imaging method based on phase recovery, irradiate the imaging samples through an X-ray light source, and use a detector to measure the phase change information in the diffraction pattern to build a phase analysis channel, including phase recovery branches and image reconstruction branches, which are carried out through a deep learning architecture. Phase denoising and filtering process to generate target diffraction images.

Benefits of technology

The contrast and resolution of imaging are improved, and the internal structure information of the substance can be obtained more clearly, solving the problem of poor resolution of traditional X-ray imaging.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118566275B_ABST
    Figure CN118566275B_ABST
Patent Text Reader

Abstract

The present invention discloses a diffraction imaging method based on phase recovery, which relates to the field of optical elements. The method comprises: using an X-ray light source to irradiate a sample to be imaged to obtain multiple diffraction patterns, performing phase measurement on multiple diffraction patterns through a detector to obtain phase change information, constructing a phase analysis channel, synchronizing the phase change information to a phase recovery branch, obtaining phase recovery information, mapping the phase recovery information to an image reconstruction branch, generating a target diffraction image, connecting a remote display, and transmitting the target diffraction image to the remote display for imaging display. The method solves the technical problem that traditional X-ray imaging mainly relies on the intensity change after the X-ray penetrates the object to form an image, resulting in poor imaging resolution, and achieves the technical effect of obtaining internal structure information by using the phase change after the X-ray passes through the substance, thereby improving the contrast and resolution of the imaging.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of optical elements, and in particular to a diffraction imaging method based on phase recovery. Background Art

[0002] With the advancement of science and technology, especially in the fields of materials science, biomedicine and physics research, the exploration of the microstructure of matter has become more and more in-depth. Traditional X-ray imaging technology mainly relies on the difference in the absorption of X-rays by materials to form images, but in some cases, this difference is not obvious and it is difficult to provide sufficient information. Traditional X-ray imaging mainly relies on the intensity change after the X-ray penetrates the object to form an image, resulting in the technical problem of poor imaging resolution. Summary of the invention

[0003] The present application provides a diffraction imaging method based on phase recovery, which solves the technical problem that traditional X-ray imaging mainly relies on the intensity change after the X-ray penetrates the object to form an image, resulting in poor imaging resolution. It achieves the technical effect of utilizing the phase change after the X-ray passes through the material to obtain internal structure information and improve the contrast and resolution of the imaging.

[0004] The present application provides a diffraction imaging method based on phase recovery, comprising: irradiating a sample to be imaged with an X-ray light source to obtain a plurality of diffraction patterns; performing phase measurement on the plurality of diffraction patterns through a detector, and recording the phase measurement results to obtain phase change information; constructing a phase analysis channel, wherein the phase analysis channel comprises a phase recovery branch and an image reconstruction branch; synchronizing the phase change information to the phase recovery branch to obtain phase recovery information, mapping the phase recovery information to the image reconstruction branch to generate a target diffraction image; and connecting a remote display to transmit the target diffraction image to the remote display for imaging display.

[0005] In a possible implementation, an X-ray light source is used to irradiate a sample to be imaged, and the following processing is performed: a plurality of irradiation detection points are determined based on the irradiation position of the X-ray light source; a plurality of detectors are arranged according to the plurality of irradiation detection points; the X-ray light source is controlled to irradiate the sample to be imaged, and the plurality of detectors receive the X-ray light irradiating the sample to be imaged to generate a plurality of patterns to be imaged, wherein the plurality of patterns to be imaged include a plurality of angular diffraction patterns and a plurality of point position diffraction images; and the plurality of angular diffraction patterns and the plurality of point position diffraction images are integrated to obtain the plurality of diffraction patterns.

[0006] In a possible implementation, the phase of the multiple diffraction patterns is measured by a detector, and the following processing is performed: the multiple diffraction patterns are measured by using the detector to determine the light intensity distribution space of the multiple diffraction patterns; the light intensity distribution space of the multiple diffraction patterns is calculated according to the multiple point diffraction patterns to generate multiple phase distribution information, and the multiple phase distribution information has a corresponding relationship with the multiple diffraction patterns; phase recording is performed based on the multiple phase distribution information to obtain the phase measurement result.

[0007] In a possible implementation, the following processing is performed: the detector has multiple pixel units, the multiple pixel units detect and record the light intensity distribution data of the multiple diffraction patterns in space, the light intensity value detected by each pixel unit is mapped to the corresponding spatial position in the light intensity distribution data, and the light intensity distribution space of the multiple diffraction patterns is determined.

[0008] In a possible implementation, the phase change information is synchronized to the phase recovery branch to obtain phase recovery information, and the following processing is performed: the phase change information is pre-processed by a data processing unit in the phase recovery branch to generate a phase change trend; the phase is initialized based on the phase change trend to generate multiple iteration parameters, and the multiple iteration parameters include a phase initial value, an iteration step, and a preset number of iterations; according to the iteration step, the phase initial value is optimized for the first iteration to generate first phase change information, and the first phase change information includes a first phase value and a first update amount; the first phase value is updated using the first update amount to determine whether the change in the i-th phase value is less than a preset threshold value, and if so, the i-th phase value is used as the target phase information and the phase recovery information is output; otherwise, the iterative calculation is continued until the preset number of iterations is met, and the n-th phase value is used as the target phase information and the phase recovery information is output, where i is any phase value within the preset number of iterations, and n is the number of preset number of iterations.

[0009] In a possible implementation, the phase change information is preprocessed by a data processing unit within the phase recovery branch to generate a phase change trend, and the following processing is performed: the phase change information is denoised using a denoising network to generate phase denoising information; the phase denoising information is filtered by a bandpass filter to generate phase denoising filter information; the phase denoising filter information is calibrated to generate a phase change trend based on the calibration result.

[0010] In a possible implementation, the phase change information is denoised using a denoising network to generate phase denoising information, and the following processing is performed: a deep learning architecture of the denoising network is constructed based on a recurrent neural network; based on the deep learning architecture, a plurality of subnetworks are determined, and the plurality of subnetworks may include a feature extraction subnetwork, a noise prediction subnetwork, and a denoising subnetwork; the phase change information is used as input data, and a first phase change feature is obtained through the feature extraction subnetwork, and the first phase change feature has a one-to-one correspondence with the low-noise phase change information; the low-noise phase change information is used as a label, and the phase change information is supervisedly trained through the noise prediction subnetwork, and the training result is denoised and verified based on the denoising subnetwork, and the phase denoising information is output.

[0011] In a possible implementation, the phase recovery information is mapped to the image reconstruction branch to generate a target diffraction image, and the following processing is performed: data recovery is performed on the multiple diffraction patterns based on the phase recovery information to generate complex domain data; the complex domain data is mapped to the image reconstruction branch to convert the complex domain data from the spatial domain to the frequency domain; filtering is performed on the complex domain data based on the frequency domain, and the frequency domain data is converted to the spatial domain according to the filtering result to generate a reconstructed image, and the target diffraction image is output.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0013] The diffraction imaging method based on phase recovery provided in the present application solves the technical problem that traditional X-ray imaging mainly relies on the intensity change after the X-ray penetrates the object to form an image, resulting in poor imaging resolution. It achieves the technical effect of utilizing the phase change after the X-ray passes through the material to obtain internal structure information, thereby improving the contrast and resolution of the imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the method according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic diagram of a process flow of a diffraction imaging method based on phase retrieval provided in an embodiment of the present application;

[0016] Figure 2A schematic diagram of the process of obtaining phase recovery information in the phase recovery based diffraction imaging method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] The present application embodiment provides a diffraction imaging method based on phase retrieval, such as Figure 1 As shown, the method includes:

[0021] Step A100, irradiating the sample to be imaged with an X-ray light source to obtain a plurality of diffraction patterns;

[0022] In a possible implementation, step A100 further includes step A110, determining a plurality of irradiation detection points based on the irradiation position of the X-ray light source; executing step A120, arranging a plurality of detectors according to the plurality of irradiation detection points;

[0023] The initial direction of the X-ray is determined based on the exact position of the X-ray light source, and multiple positions of the diffraction pattern of the X-ray are determined according to the size, shape and required imaging accuracy of the sample to be imaged, thereby determining multiple irradiation detection points, which can be distributed around the sample to be imaged to obtain diffraction information at different angles, and then one or more detectors are installed at each predetermined irradiation detection point. These detectors can be flat panel detectors, CCD cameras or other devices suitable for receiving X-rays, and ensure that all detectors are properly installed and calibrated so that they can accurately receive and record the diffraction pattern of the X-rays.

[0024] Execute step A130, control the X-ray light source to irradiate the sample to be imaged, receive the X-ray light irradiating the sample to be imaged through the multiple detectors, and generate multiple patterns to be imaged, wherein the multiple patterns to be imaged include multiple angle diffraction patterns and multiple point diffraction images; execute step A140, integrate the multiple angle diffraction patterns and the multiple point diffraction images to obtain the multiple diffraction patterns.

[0025] Controlling the X-ray light source to irradiate the sample to be imaged means starting the X-ray light source and controlling it to irradiate the sample to be imaged with predetermined parameters (such as power, irradiation time, etc.). When the X-rays are irradiated onto the sample to be imaged, the X-rays will interact with the atoms or molecules in the sample to be imaged to produce diffraction phenomena. The diffracted X-rays are then received by the arranged detectors, and each detector will record multiple to-be-imaged patterns of the received X-rays. The multiple to-be-imaged patterns include diffraction patterns at multiple angles and diffraction images at multiple points. Finally, the recorded diffraction pattern data is collected from all detectors, and these diffraction patterns from different angles and points are integrated. The integration process may include steps such as alignment, superposition, and filtering to eliminate noise, enhance signals, and improve imaging quality. After integration processing, multiple diffraction patterns are merged into one or more final imaging results. These imaging results may include two-dimensional images, three-dimensional models, or other forms of data representation, so as to obtain multiple diffraction patterns.

[0026] Execute step A200, perform phase measurement on the multiple diffraction patterns by using a detector, and record the phase measurement results to obtain phase change information;

[0027] In a possible implementation, step A200 further includes step A210, measuring the multiple diffraction patterns using the detector to determine the light intensity distribution space of the multiple diffraction patterns;

[0028] Measuring the multiple diffraction patterns using the detector means that the intensity of the X-rays at each position in each diffraction pattern can be measured by the detector, thereby forming a light intensity distribution space, and the light intensity distribution space can be a two-dimensional image, in which the value of each pixel represents the X-ray intensity at that position. Furthermore, the detector is a grating-based interferometer or pixel array detector, which can capture the interference effect generated by the propagation of X-rays inside the material, thereby accurately measuring slight changes in the phase. The detector is an important device for capturing diffraction patterns, and its performance directly affects the accuracy and quality of imaging. The detector has multiple pixel units, and the multiple pixel units detect and record the light intensity distribution data of the multiple diffraction patterns in space. Each pixel unit is a photosensitive element that can detect the intensity of the X-rays irradiated thereon. After the X-ray passes through the target object, due to the diffraction effect, a specific light intensity distribution will be formed in space, so that the light intensity value detected by each pixel unit is mapped to the corresponding spatial position in the light intensity distribution data, which means that the light intensity value is converted into an electrical signal based on the intensity of the X-rays sensed by multiple pixel units. In addition to detecting the light intensity, the multiple pixel units also have the function of spatial positioning. In the detector, each pixel unit has a fixed position coordinate. When the X-ray is irradiated on the detector, each pixel unit can accurately know the position of its pixel, so that the detected light intensity value can be mapped to the corresponding spatial position. When the X-ray passes through the target object, the X-ray light intensity distribution in space will be formed on the detector, thereby determining the light intensity distribution space of multiple diffraction patterns.

[0029] Execute step A220, calculate the light intensity distribution space of the multiple diffraction patterns according to the multiple point diffraction patterns, and generate multiple phase distribution information, and the multiple phase distribution information has a corresponding relationship with the multiple diffraction patterns; execute step A230, perform phase recording based on the multiple phase distribution information, and obtain the phase measurement result.

[0030] Since only intensity information can be directly measured by the detector during the X-ray diffraction process, while phase information will be lost, a phase recovery algorithm (such as iterative Fourier transform algorithm, Gerchberg-Saxton algorithm, etc.) is needed to recover phase information from the light intensity distribution. The phase recovery algorithm can be used to calculate the phase distribution corresponding to each diffraction pattern. The phase distribution can be a two-dimensional image, in which the value of each pixel represents the X-ray phase at that position. Since there are multiple point diffraction patterns, multiple corresponding phase distribution information will be generated. The multiple phase distribution information has a one-to-one correspondence with the diffraction pattern. The multiple phase distribution information calculated is recorded as a phase measurement result. At the same time, if multiple measurements are performed, for example, at different times, under different conditions or for different samples, all the obtained phase measurement results can be compared to obtain phase change information. The phase change information can be used to characterize the dynamic behavior, structural changes or other physical phenomena of the sample.

[0031] Execute step A300 to construct a phase analysis channel, where the phase analysis channel includes a phase recovery branch and an image reconstruction branch;

[0032] The construction of the phase analysis channel involves two key branches: phase recovery and image reconstruction. The phase recovery branch recovers the phase distribution information through the collected diffraction pattern, while the image reconstruction branch reconstructs the visual image using this phase distribution information and the original data.

[0033] Furthermore, the above-mentioned multiple detectors can receive X-ray light irradiating the sample to be imaged based on the preset multiple irradiation detection points to generate multiple patterns to be imaged, and the image to be imaged includes multiple angle diffraction patterns and multiple point diffraction images, and the phase recovery algorithm is determined according to the characteristics of the sample to be imaged and the imaging requirements, and the collected multiple diffraction patterns are used as input data, and the selected phase recovery algorithm is applied for calculation, and the phase information is recovered from the light intensity distribution through an iterative and optimization process, and the sample to be imaged is phase recovered after algorithm processing, and multiple phase recovery information corresponding to the diffraction patterns is generated. The phase recovery information is used to reflect the internal structure and characteristics of the sample to be imaged, thereby completing the construction of the phase recovery branch.

[0034] Furthermore, the phase recovery information generated by the phase recovery branch and the original diffraction pattern data are integrated, and an image reconstruction algorithm is selected according to the imaging requirements and sample characteristics. The image reconstruction algorithm may include a filtered back projection algorithm, an algebraic iteration method, etc., and the phase recovery information and the diffraction pattern data are calculated and optimized according to the selected image reconstruction algorithm, and the phase recovery information and the diffraction pattern data are converted into a visual image. The reconstructed image is subjected to necessary post-processing operations, such as smoothing and enhancement, to improve the image quality and readability, thereby completing the construction of the image reconstruction branch, by using the input end of the phase recovery branch as the input end of the phase analysis channel, and the output end of the phase recovery branch is connected to the input end of the image reconstruction branch for communication, and the output end of the image reconstruction branch is used as the output end of the phase analysis channel.

[0035] Execute step A400, synchronize the phase change information to the phase recovery branch to obtain phase recovery information, map the phase recovery information to the image reconstruction branch, and generate a target diffraction image;

[0036] In one possible implementation, Figure 2 As shown, step A400 further includes step A410, preprocessing the phase change information by a data processing unit in the phase recovery branch to generate a phase change trend; in a possible implementation, step A410 further includes step A411, denoising the phase change information by using a denoising network to generate phase denoising information;

[0037] In one possible implementation, step A411 further includes step A4111, constructing a deep learning architecture of the denoising network based on a recurrent neural network; executing step A4112, determining multiple subnetworks according to the deep learning architecture, and the multiple subnetworks may include a feature extraction subnetwork, a noise prediction subnetwork, and a denoising subnetwork; executing step A4113, taking the phase change information as input data, and obtaining a first phase change feature through the feature extraction subnetwork, and the first phase change feature has a one-to-one correspondence with the low-noise phase change information; executing step A4114, taking the low-noise phase change information as a label, performing supervised training on the phase change information through the noise prediction subnetwork, and performing denoising verification on the training results based on the denoising subnetwork, and outputting the phase denoising information.

[0038] A deep learning architecture for processing phase change information is determined based on a recurrent neural network. The deep learning architecture is a neural network structure including multiple levels, and a feature extraction subnetwork, a noise prediction subnetwork, and a denoising subnetwork are determined in the deep learning architecture. The feature extraction subnetwork is used to extract key features from the input phase change information, the noise prediction subnetwork uses the extracted features to predict the noise in the phase change information, and the denoising subnetwork is used to remove the predicted noise to obtain denoised phase change information.

[0039] Furthermore, the phase change information is used as input data, and the input data may be time series data, image data, or data in other formats. Depending on the representation of the phase change information, the real value may be obtained by physical methods or other means, or the ideal value may be generated by simulation, to obtain low-noise phase change information, and the low-noise phase change information is used as label data. At the same time, the phase change information is input into the feature extraction subnetwork, and the feature extraction subnetwork extracts key feature information from the input data according to multi-layer convolution, pooling, and other operations to generate a first phase change feature. These features usually have a one-to-one correspondence with the low-noise phase change information, and then the low-noise phase change information is input into the noise prediction subnetwork as label data. The noise prediction subnetwork uses the feature information extracted from the feature extraction subnetwork to predict the noise in the input phase change information, and by comparing the predicted noise with the real noise (i.e., the difference between the label data and the input data), the weights and bias parameters of the network are updated using the back propagation algorithm and the gradient descent method to perform supervised training.

[0040] Furthermore, the predicted noise is input into the denoising subnetwork, and the noise is removed by a denoising algorithm (such as residual learning, gating mechanism, etc.) to obtain denoised phase change information, which is then verified using a validation set or a test set. If the preset performance indicators are met, which may include signal-to-noise ratio, structural similarity, etc., the phase denoising information is output, otherwise, the network structure or parameters may be adjusted, and retraining and verification may be performed. During the training process, the network parameters and structure may be continuously optimized in an iterative manner to improve the denoising performance.

[0041] Execute step A412, filter the phase denoising information by bandpass filtering to generate phase denoising filter information; execute step A413, calibrate the phase denoising filter information, and generate a phase change trend according to the calibration result.

[0042] The phase denoising information is filtered by bandpass filtering, and the bandpass filtering is used to retain the components within a specific frequency range in the phase denoising information, while suppressing or eliminating noise and interference in other frequency ranges. First, the frequency range of the bandpass filter is determined according to the characteristics of the phase denoising information. Secondly, based on the selected frequency range, the type of bandpass filter, such as Butterworth, Chebyshev, etc., and the order are determined. The phase denoising information is further input into the designed bandpass filter to obtain the filtered phase denoising information, that is, the phase denoising filter information.

[0043] Calibrating the phase denoising filter information refers to further eliminating the system error or deviation in the phase denoising filter information, and may include some linear or nonlinear transformations to better reflect the real trend of the phase change. According to the characteristics of the phase denoising filter information, the parameters that need to be calibrated are determined. The calibration parameters may include offset, gain, linearity, etc. Based on the selected calibration parameters, the calibration algorithm is determined. The calibration algorithm may involve linear transformation, polynomial fitting, lookup table and other methods, and the phase denoising filter information is input into the calibration algorithm. The calibrated phase information is calculated according to the algorithm. The trend of phase change is extracted through the calibrated phase information, which refers to time domain or frequency domain analysis of the calibrated phase information, recording and analyzing its changes over time or frequency, and extracting the trend of phase change according to the analysis results. It can include the slope, period, amplitude and other parameters of the calculated phase change, so as to output the trend of phase change.

[0044] Execute step A420, perform phase initialization based on the phase change trend, and generate multiple iteration parameters, wherein the multiple iteration parameters include an initial phase value, an iteration step, and a preset number of iterations;

[0045] Based on the phase change trend, a suitable initial phase value is set, and the relevant parameters of iterative optimization are determined. According to the phase change trend, a phase value is selected as the initial phase value. The selected initial value is the closest to the actual phase value to improve the efficiency of iterative optimization. The iteration step size refers to the update amount of the phase value in each iteration, and it can also be weighed based on the stability and convergence speed of the system. The preset number of iterations refers to the upper limit of the number of iterations to ensure that the algorithm converges to the optimal solution within a reasonable time.

[0046] Execute step A430, perform a first iterative optimization on the initial phase value according to the iteration step, and generate first phase change information, wherein the first phase change information includes a first phase value and a first update amount; execute step A440, use the first update amount to update the first phase value, and determine whether the change of the i-th phase value is less than a preset threshold value. If so, use the i-th phase value as the target phase information and output the phase recovery information. Otherwise, continue to perform iterative calculation until the preset number of iterations is met, use the n-th phase value as the target phase information and output the phase recovery information, where i is any phase value within the preset number of iterations, and n is the number of preset number of iterations.

[0047] According to the iteration step, the first iteration optimization of the initial phase value means that a new phase change information including the updated phase value and the update amount is obtained through the first iteration optimization, that is, according to the set iteration step, the initial phase value is updated to obtain the phase value after the first iteration, and the update amount of the phase value is calculated during the iteration. The calculated update amount can be used for subsequent phase value updates and convergence judgments. Further, through multiple iteration calculations, the real phase value is gradually approached, and the iteration is stopped when the convergence condition is met. It means that the current phase value is updated using the update amount calculated in the previous iteration, and whether the change amount of the current phase value is less than the preset threshold value is judged. If the change amount of the current phase value is less than the preset threshold value, it is considered that the algorithm has converged to the optimal solution, and the current phase value, that is, the i-th phase value, is output as the target phase information. Otherwise, the next iteration calculation is continued, and after each iteration, it is checked whether the preset upper limit of the number of iterations has been reached. If the upper limit is reached but the algorithm has not converged, the phase value obtained in the last iteration, that is, the n-th phase value, is output as the target phase information.

[0048] In a possible implementation, step A400 further includes step A450, performing data recovery on the multiple diffraction patterns based on the phase recovery information to generate complex domain data; executing step A460, converting the complex domain data from the spatial domain to the frequency domain by mapping the complex domain data to the image reconstruction branch;

[0049] Using phase recovery information to recover the diffraction pattern to obtain complex domain data means combining the phase recovery information with the amplitude information of the diffraction pattern to generate complex domain data, and the complex domain data contains amplitude and phase information, which together describe the complete information of the diffraction pattern.

[0050] Further conversion of complex domain data from the spatial domain to the frequency domain for further filtering processing refers to performing a two-dimensional Fourier transform (or fast Fourier transform, FFT) on the complex domain data to convert it from the spatial domain to the frequency domain, which is achieved through Fourier transform. Fourier transform is a mathematical tool that can convert complex domain data from the spatial domain to the frequency domain. In complex domain data, each data point contains amplitude and phase information. Fourier transform analyzes the spatial distribution of this data and outputs the corresponding frequency components.

[0051] Execute step A470, perform a filtering operation on the complex domain data based on the frequency domain, convert the frequency domain data to the spatial domain according to the filtering result, generate a reconstructed image, and output the target diffraction image.

[0052] Filtering the complex domain data in the frequency domain to remove noise and unnecessary frequency components refers to selecting and determining filters such as low-pass filters and band-pass filters according to the needs of image reconstruction and the characteristics of the data, and applying the filters to the frequency domain data to remove unnecessary frequency components. The filtering operation can be achieved by applying filters, which are mathematical functions that can perform weighted processing on the spectrum of the image. Different types of filters can achieve different effects. For example, a low-pass filter can remove high-frequency noise, while a high-pass filter can enhance the details and edges of the image. A band-pass filter can selectively retain components within a certain frequency range, and then convert the filtered frequency domain data back to the spatial domain to generate a reconstructed image. In the frequency domain, different frequency components of the image correspond to different visual features. For example, low-frequency components usually correspond to smooth areas of the image, while high-frequency components correspond to details and edges of the image. Therefore, filtering the complex domain data in the frequency domain can selectively remove noise or enhance specific image features.

[0053] In order to convert it back to the spatial domain and generate the final image, an inverse Fourier transform is needed, which refers to performing a two-dimensional inverse Fourier transform or an inverse fast Fourier transform on the filtered frequency domain data to convert it from the frequency domain back to the spatial domain. The inverse Fourier transform is the inverse process of the Fourier transform. The inverse Fourier transform can convert the spectral representation of the image back to the spatial domain representation, and visualize the spatial domain data obtained by the inverse Fourier transform as an image, that is, to obtain a reconstructed image. Finally, the reconstructed image is output as a target diffraction image for subsequent analysis or application.

[0054] Next, step A500 is performed to connect to a remote display and transmit the target diffraction image to the remote display for imaging display.

[0055] To connect to a remote display and transmit the target diffraction image to it for imaging display, you can first ensure that the remote display and the target computer are connected through a stable network, turn on the remote connection function on the computer where the target diffraction image is located, install and start the selected remote connection tool on the computer where the remote display is located, and then use the remote desktop protocol or professional remote connection software to start a remote connection session on the target computer, and share the target diffraction image file or display window to the remote display. Once the connection is successful, the remote display can display the target diffraction image in real time, which means that the remote display will display the target diffraction image transmitted from the computer where the target diffraction image is located. According to the settings of the display and the properties of the image file, the image may be displayed at an appropriate size and resolution, so as to transmit the target diffraction image to the remote display for imaging display, and realize remote real-time image viewing and operation.

[0056] The embodiments of the present application solve the technical problem that traditional X-ray imaging mainly relies on the intensity change after the X-ray penetrates the object to form an image, resulting in poor imaging resolution. The embodiments of the present application achieve the technical effect of utilizing the phase change after the X-ray passes through the material to obtain internal structure information, thereby improving the contrast and resolution of the imaging.

[0057] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A diffraction imaging method based on phase retrieval, characterized in that: The method comprises: Using an X-ray light source to irradiate the sample to be imaged to obtain multiple diffraction patterns; Performing phase measurement on the multiple diffraction patterns through a detector, and recording the phase measurement results to obtain phase change information; Constructing a phase analysis channel, wherein the phase analysis channel comprises a phase recovery branch and an image reconstruction branch; Synchronizing the phase change information to the phase recovery branch to obtain phase recovery information, and mapping the phase recovery information to the image reconstruction branch to generate a target diffraction image; Connecting to a remote display, transmitting the target diffraction image to the remote display for imaging display; Preprocessing the phase change information by a data processing unit in the phase recovery branch to generate a phase change trend; Performing phase initialization based on the phase change trend to generate multiple iteration parameters, wherein the multiple iteration parameters include an initial phase value, an iteration step, and a preset number of iterations; According to the iteration step, performing a first iteration optimization on the initial phase value to generate first phase change information, wherein the first phase change information includes a first phase value and a first update amount; Using the first update amount to update the first phase value, determine whether the change of the i-th phase value is less than a preset threshold value, if so, use the i-th phase value as the target phase information, and output the phase recovery information, if not, continue to perform iterative calculation until the preset number of iterations is met, use the n-th phase value as the target phase information, and output the phase recovery information, where i is any phase value within the preset number of iterations, and n is the number of preset number of iterations; Performing denoising processing on the phase change information using a denoising network to generate phase denoising information; Filtering the phase denoising information by bandpass filtering to generate phase denoising filtering information; Calibrate the phase denoising filter information, and generate a phase change trend according to the calibration result; Constructing a deep learning architecture of the denoising network based on a recurrent neural network; According to the deep learning architecture, determining a plurality of sub-networks, wherein the plurality of sub-networks may include a feature extraction sub-network, a noise prediction sub-network, and a denoising sub-network; Taking the phase change information as input data, obtaining a first phase change feature through the feature extraction subnetwork, wherein the first phase change feature has a one-to-one correspondence with the low-noise phase change information; Using the low-noise phase change information as a label, performing supervised training on the phase change information through the noise prediction subnetwork, performing denoising verification on the training result based on the denoising subnetwork, and outputting the phase denoising information; Performing data recovery on the plurality of diffraction patterns based on the phase recovery information to generate complex domain data; By mapping the complex domain data to the image reconstruction branch, the complex domain data is converted from the spatial domain to the frequency domain; A filtering operation is performed on the complex domain data based on the frequency domain, and the frequency domain data is converted to the spatial domain according to the filtering result to generate a reconstructed image, and the target diffraction image is output.

2. The diffraction imaging method based on phase retrieval according to claim 1, characterized in that: The sample to be imaged is irradiated with an X-ray light source to obtain a plurality of diffraction patterns, the method comprising: Determining a plurality of irradiation detection points based on the irradiation position of the X-ray light source; Arrange a plurality of detectors according to the plurality of irradiation detection points; Controlling the X-ray light source to irradiate the sample to be imaged, and receiving the X-ray light irradiating the sample to be imaged through the multiple detectors to generate multiple patterns to be imaged, wherein the multiple patterns to be imaged include multiple angle diffraction patterns and multiple point diffraction images; The multiple angle diffraction patterns and the multiple point position diffraction images are integrated to obtain the multiple diffraction patterns.

3. The diffraction imaging method based on phase retrieval according to claim 2, characterized in that: Performing phase measurement on the multiple diffraction patterns by a detector, the method comprising: Measuring the multiple diffraction patterns using the detector to determine the light intensity distribution space of the multiple diffraction patterns; Calculating the light intensity distribution space of the multiple diffraction patterns according to the multiple point diffraction patterns to generate multiple phase distribution information, wherein the multiple phase distribution information has a corresponding relationship with the multiple diffraction patterns; Phase recording is performed based on the multiple phase distribution information to obtain the phase measurement result.

4. The diffraction imaging method based on phase retrieval according to claim 3, characterized in that: The detector has a plurality of pixel units, which detect and record the light intensity distribution data of the plurality of diffraction patterns in space, map the light intensity value detected by each pixel unit to the corresponding spatial position in the light intensity distribution data, and determine the light intensity distribution space of the plurality of diffraction patterns.