X-ray grating comprehensive imaging information extraction method and system

Abstract

The invention relates to the technical field of X-ray imaging, and discloses an X-ray grating comprehensive imaging information extraction method and system. The method comprises the steps of obtaining and preprocessing original data; acquiring multi-dimensional interference parameters formed by grating phase drift distance, photon statistical noise density and imaging object deformation rate; dynamically calculating a phase correction coefficient, a noise suppression weight and information extraction confidence; phase compensation and noise filtering are executed based on the confidence coefficient, and structure and density information is extracted; parameter changes are monitored in real time, and the extraction process is updated; the system correspondingly realizes the method. The problem of low dynamic scene information precision caused by multi-interference coupling is solved, and the imaging information extraction stability and quality are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of X-ray imaging technology, in particular to an X-ray grating comprehensive imaging information extraction method and system. BACKGROUND

[0002] X-ray grating imaging technology is widely used in medical diagnosis, such as lung dynamic ventilation imaging, cardiovascular motion imaging, and industrial detection, such as internal defect detection of high-speed rotating parts, due to its high spatial resolution and high soft tissue contrast. The core of this technology is to modulate the X-ray beam through the grating, and then extract the structure information and density information of the imaging object based on the modulated data collected by the detector.

[0003] The existing X-ray grating imaging information extraction method usually only corrects for a single interference factor when processing imaging data, without considering the coupling effects of grating phase drift, photon statistical noise and imaging object deformation. In dynamic imaging scenarios, the grating phase drifts due to environmental temperature fluctuations and mechanical vibrations, resulting in a decrease in X-ray modulation accuracy; photon statistical noise fluctuates with the dynamic changes in X-ray radiation dose, and noise interference is particularly significant in low-dose scenarios; the slight deformation of the imaging object, such as lung movement caused by respiration and vibration during industrial component rotation, further adds to the interference, and the interaction of the three factors results in severe distortion of the original data. This multi-interference coupling problem causes the extracted structure information to be blurred and the density information to have artifacts, which cannot meet the high-precision requirements of medical diagnosis for subtle structure identification and industrial detection for defect precise positioning.

[0004] Based on the above problems, there is an urgent need for a technical solution that can handle multiple interference coupling problems simultaneously and improve the information extraction accuracy and stability in dynamic scenarios. SUMMARY

[0005] The purpose of the present application is to solve the problems existing in the prior art and to provide an X-ray grating comprehensive imaging information extraction method, which includes acquiring original data of X-ray grating imaging and preprocessing the original data; collecting multi-dimensional interference parameters, including grating phase drift, photon statistical noise density and imaging object deformation rate; based on the multi-dimensional interference parameters, dynamically calculating phase correction coefficients, noise suppression weights and information extraction confidence; determining whether the information extraction confidence meets the pre-set confidence threshold, if yes, performing phase compensation on the preprocessed original data based on the phase correction coefficients, and then performing noise filtering in combination with the noise suppression weights, and finally extracting the structure information and density information of the imaging object; real-time monitoring the change state of the multi-dimensional interference parameters, if the change amount of any parameter in the multi-dimensional interference parameters triggers the update condition, then recalculating the phase correction coefficients, noise suppression weights and information extraction confidence, and updating the information extraction process according to the new calculation results.

[0006] Preferably, in the step of collecting multi-dimensional interference parameters, the grating phase shift is collected by a laser interferometer, the sampling frequency of the laser interferometer matches the period of the X-ray grating, and the collected data is subjected to Fourier transform to obtain the amount of change of the grating phase over time; the photon statistical noise density is obtained by calculating the pixel gray scale variance of the original data after continuous multiple frames of preprocessing, and a background region without an imaging object in the imaging region is selected as a sample region during calculation; the imaging object deformation rate is obtained by a feature point matching algorithm, the feature points are extracted in the current frame of imaging data, and then the corresponding feature points in the previous frame of imaging data are matched, the ratio of the feature point displacement to the interval time between the two frames is calculated, and the average of the ratios of all feature points is taken as the imaging object deformation rate.

[0007] Preferably, the step of preprocessing the original data includes dark current correction, bad pixel repair, and gray scale normalization. The dark current correction is achieved by subtracting the average of the dark current data output by the detector when there is no X-ray radiation from the original data; the bad pixel repair uses a neighborhood interpolation method, and the gray scale value of a pixel deviating from the average of the neighborhood by more than a preset deviation value is replaced by the average of the gray scale values of the neighborhood pixels; and the gray scale normalization maps the gray scale values of the original data to the range of 0-255.

[0008] Preferably, in the step of judging whether the information extraction confidence meets a preset confidence threshold, the preset confidence threshold is set based on the imaging scene requirements, and the preset confidence threshold in the medical dynamic organ imaging scene is higher than that in the industrial static component detection scene; if the information extraction confidence does not meet the preset confidence threshold, the collection frequency of the multi-dimensional interference parameters is adjusted, the multi-dimensional interference parameters are re-collected after increasing the collection frequency, and the phase correction coefficient, the noise suppression weight, and the information extraction confidence are calculated until the information extraction confidence meets the preset confidence threshold.

[0009] Preferably, the formula for dynamically calculating the phase correction coefficient is:

[0010] ;

[0011] wherein, is the phase correction coefficient, dimensionless; is the grating phase shift, with the unit of rad; is the imaging object deformation rate, with the unit of mm / s; is the reference deformation rate, which is the maximum allowed deformation rate adapted by the imaging system; is the phase compensation adjustment factor, with the value ranging from 0.1 to 0.3; is the phase shift reference value, with the unit of rad, which is the maximum phase fluctuation amount when the grating is stably working; is a natural exponential function with base number about 2.718; is a cosine function with input parameter unit rad.

[0012] Preferably, the formula for dynamically calculating the noise suppression weight is:

[0013] wherein, is the noise suppression weight, dimensionless; is the photon statistical noise density, unit photon / pixel; is the maximum allowed noise density, unit photon / pixel; is the phase correction coefficient defined above; is the noise adjustment factor, with a value range of 0.2 to 0.4; is the signal-to-noise ratio of the imaging data, calculated from the pre-processed original data, and the calculation method is the ratio of the mean gray value of the signal region to the standard deviation of the gray value of the background region; is a logarithmic function with base number 10.

[0014] Preferably, the formula for dynamically calculating the information extraction confidence is:

[0015] ;

[0016] wherein, is the information extraction confidence, dimensionless, with a value range of 0 to 1; is the noise suppression weight defined above; is the phase correction coefficient defined above; is the reference phase correction coefficient, with a value of 1.0; is the confidence adjustment factor, with a value range of 0.15 to 0.25; is the mean gray value of the current frame of imaging data; is the mean gray value of the last 10 frames of imaging data; is a natural exponential function with base number about 2.718.

[0017] Preferably, the X-ray grating comprehensive imaging information extraction system is applied to the X-ray grating comprehensive imaging information extraction method as described in any one of the above, and comprises a data acquisition module, an interference parameter acquisition module, a calculation module, an information extraction module and a monitoring and updating module. The data acquisition module comprises an X-ray source, a grating assembly and a detector array. The X-ray source is used to emit an X-ray beam. The grating assembly is arranged between the X-ray source and an imaging object. The detector array is used to receive X-rays that penetrate the imaging object and the grating assembly and output original data of X-ray grating imaging. The interference parameter acquisition module comprises a laser interferometer, a noise analysis unit and a deformation detection unit. The laser interferometer is used to acquire a grating phase drift amount. The noise analysis unit is used to calculate a photon statistical noise density. The deformation detection unit is used to acquire an imaging object deformation rate. The calculation module is used to dynamically calculate a phase correction coefficient, a noise suppression weight and an information extraction confidence level based on multi-dimensional interference parameters. The information extraction module is used to judge whether the information extraction confidence level meets a preset threshold value, and to perform phase compensation, noise filtering and information extraction. The monitoring and updating module is used to monitor a change state of the multi-dimensional interference parameters in real time, and to control the calculation module to recalculate and update the information extraction process when a trigger update condition is met.

[0018] Preferably, the X-ray source in the data acquisition module adopts a micro-focus X-ray source. The emitted X-ray beam energy is adjustable, and the adjustment range is set according to the material and thickness of the imaging object. The grating assembly comprises a phase grating and an absorption grating. The periods of the phase grating and the absorption grating are matched. The phase grating is used to modulate the phase of the X-ray beam. The absorption grating is used to convert the phase modulation into intensity modulation. The detector array adopts a flat panel detector. The pixel size of the detector array is adapted to the grating period. The detector array is connected to the calculation module through a data transmission interface. The data transmission interface adopts a CameraLink interface. The transmission rate meets the real-time imaging requirement.

[0019] Preferably, the monitoring and updating module comprises a parameter monitoring unit and an update trigger unit. The parameter monitoring unit acquires a current value of the multi-dimensional interference parameters at a preset period, and calculates a change amount of the current value and a value acquired in a previous acquisition period. The update trigger unit pre-stores an update condition. The update condition is that the change amount of any parameter in the multi-dimensional interference parameters exceeds 15% of a near-10-minute historical average value. If the parameter change amount meets the update condition, the update trigger unit sends a recalculation instruction to the calculation module, and controls the information extraction module to pause the current information extraction process. After the calculation module completes the recalculation, the information extraction module restarts the information extraction process according to the new calculation result.

[0020] Technical effects: The present application collects multi-dimensional interference parameters composed of grating phase shift, photon statistical noise density and imaging object deformation rate, dynamically calculates phase correction coefficient, noise suppression weight and information extraction confidence and links to optimize, solves the problem of low dynamic scene information accuracy caused by the existing technology not processing multi-interference coupling, realizes stable extraction of imaging object structure and density information, and guarantees information quality in dynamic imaging scene. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The X-ray grating comprehensive imaging information extraction method flowchart of the present application;

[0022] Figure 2 The X-ray grating comprehensive imaging information extraction system connection block diagram of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0024] The traditional technical scheme has the following technical problems: the existing X-ray grating imaging information extraction method does not consider the coupling interference of grating phase shift, photon statistical noise and imaging object deformation, the original data is severely distorted in dynamic scene, the extracted structure information edge is blurred, the density information has artifacts, and the high-precision imaging demand cannot be met.

[0025] Based on this, please refer to Figure 1 and Figure 2 The present application provides an X-ray grating comprehensive imaging information extraction method, comprising:

[0026] S1: obtaining the original data of X-ray grating imaging and pre-processing the original data;

[0027] S2: collecting multi-dimensional interference parameters, the multi-dimensional interference parameters comprising grating phase shift, photon statistical noise density and imaging object deformation rate;

[0028] S3: based on the multi-dimensional interference parameters, dynamically calculating phase correction coefficient, noise suppression weight and information extraction confidence;

[0029] S4: judging whether the information extraction confidence meets the pre-set confidence threshold, if yes, performing phase compensation on the pre-processed original data based on the phase correction coefficient, then combining the noise suppression weight to perform noise filtering, and finally extracting the structure information and density information of the imaging object;

[0030] S5: Monitor the changes in the multi-dimensional interference parameters in real time. If the change in any of the multi-dimensional interference parameters triggers the update condition, recalculate the phase correction coefficient, noise suppression weight, and information extraction confidence level, and update the information extraction process based on the new calculation results.

[0031] The implementation of this technical solution requires a combination of hardware execution environment and software algorithm logic. In the data acquisition stage, raw data is collected using a flat panel detector with a pixel size of 50μm. A Hamamatsu C12840 flat panel detector can be used; it is adapted to the grating period to ensure complete acquisition of modulation information. In the preprocessing stage, dark current correction requires collecting dark current data 20 times without X-ray radiation, calculating the mean, and subtracting it from the raw data to eliminate the influence of detector noise. Defective pixel repair uses 3×3 neighborhood interpolation, with a preset deviation value set at 20% of the neighborhood mean; pixels exceeding the deviation are replaced. Gray-level normalization obtains the maximum and minimum gray-level values ​​by traversing the raw data, and then maps them to the 0-255 range according to a formula for subsequent calculations. When acquiring multi-dimensional interference parameters, the sampling frequency of the laser interferometer (such as ZYGOGPI) is set to 1kHz to match the grating period. The acquired light intensity signal is Fourier transformed to obtain the phase shift Δφ. The noise analysis unit selects a 200×200 pixel background area and calculates the gray-level variance of 5 consecutive frames as the photon statistical noise density ρ. noise The deformation detection unit uses the SIFT feature point extraction algorithm, extracting no less than 50 feature points per frame. After matching, the displacement is calculated. The interval between two frames is set to 0.1s to obtain the deformation rate v. def In the dynamic calculation phase, the computing module uses the Huawei Kunpeng 920 processor to simultaneously solve the three indicators through parallel computing. During information extraction, a pre-set threshold is set to 0.8 for medical scenarios and 0.7 for industrial scenarios. After meeting the threshold, phase compensation multiplies the preprocessed data by the phase correction coefficient P. Noise filtering uses bilateral filtering, and the filter kernel size is determined by W. When W=1, the kernel size is 5×5, and when W=0.5, the kernel size is 3×3. Finally, structural information is extracted through phase recovery algorithm, and density information is extracted through grayscale value inversion. In the monitoring and update phase, the parameter monitoring cycle is set to 0.5s. The historical average value in the update conditions is calculated using a sliding window algorithm based on data from the past 10 minutes. When the change exceeds 15%, recalculation is triggered to ensure that the information extraction process is dynamically adjusted according to changes in interference.

[0032] This solution addresses the problem of information distortion in dynamic scenes through multi-interference collaborative processing and real-time updates, enabling accurate extraction of structural and density information and ensuring imaging quality.

[0033] Traditional technical solutions have the following technical problems: existing multi-dimensional interference parameter acquisition methods lack unified acquisition standards and calculation logic; grating phase drift acquisition is easily affected by ambient light interference; the sample area for photon statistical noise density calculation is arbitrarily selected; and the deformation rate calculation accuracy of the imaging object is low, resulting in interference parameters failing to accurately reflect the actual interference state.

[0034] Based on this, in the step of collecting multi-dimensional interference parameters, the grating phase drift is collected by a laser interferometer, the sampling frequency of which is matched with the period of the X-ray grating, and the collected data is Fourier transformed to obtain the change in grating phase over time; the photon statistical noise density is obtained by calculating the pixel grayscale variance of the original data after preprocessing multiple consecutive frames, and the background area without imaging objects in the imaging area is selected as the sample area during the calculation; the deformation rate of the imaging object is obtained by a feature point matching algorithm, firstly extracting feature points in the current frame imaging data, then matching them with the corresponding feature points in the previous frame imaging data, calculating the ratio of the feature point displacement to the time interval between two frames, and taking the average of all feature point ratios as the deformation rate of the imaging object.

[0035] In this technical solution, the acquisition of grating phase drift requires precise alignment between the laser interferometer and the grating. The laser beam emitted by the interferometer is incident perpendicularly on the grating surface, and the interference signal is received after reflection. The sampling frequency is set according to the grating period; for example, a sampling frequency of 1kHz is used when the period is 5μm to avoid insufficient sampling leading to loss of phase information. The acquired interference signal is a curve of light intensity changing with time. The time-domain signal is converted into a frequency-domain signal through Fourier transform, and the phase component Δφ is extracted. Simultaneously, a low-pass filter is used to eliminate ambient light interference, with the filter cutoff frequency set to 10Hz. When calculating the photon statistical noise density, the background region needs to be determined through an imaging region segmentation algorithm. A threshold segmentation method is used to determine regions with gray values ​​below 20 as the background. The sample area is not less than 100×100 pixels. Five frames of data are continuously acquired, and the gray-level variance of the background region in each frame is calculated. The mean value is taken as ρ. noise To avoid errors caused by the randomness of single-frame data, the SIFT algorithm is used for feature point extraction in the calculation of the deformation rate of the imaging object. First, a scale space is constructed, extreme points are detected and low-contrast points are removed, and then edge response points are removed by principal curvature determination to ensure the stability of feature points. The K-nearest neighbor algorithm is used for feature point matching, with a matching threshold set to 0.8 to remove incorrect matching pairs. The Euclidean distance formula is used for displacement calculation, and the time interval between two frames is set according to the imaging frame rate. For example, the interval is 0.1s when the frame rate is 10fps. Finally, the average value of the rate of all effective feature points is taken. If the number of feature points is less than 20, the number of acquisition frames is increased to supplement the feature points.

[0036] This solution improves the accuracy of interference parameter collection through standardized data acquisition processes and precise calculation logic, providing reliable data support for subsequent indicator calculations and ensuring the accuracy of information extraction.

[0037] Traditional technical solutions have the following technical problems: existing raw data preprocessing methods are simple and only perform simple grayscale correction, which does not effectively eliminate the effects of dark current, bad pixels and inconsistent grayscale range. The preprocessed data still has noise residue and grayscale distortion, which affects the accuracy of subsequent information extraction.

[0038] Based on this, the preprocessing steps for the original data include dark current correction, bad pixel repair, and grayscale normalization. Dark current correction is achieved by collecting dark current data output by the detector when there is no X-ray radiation and subtracting the mean of the dark current data from the original data. Bad pixel repair uses neighborhood interpolation. For pixels whose grayscale value deviates from the mean of the neighborhood by more than a preset deviation value, the grayscale value of the pixel is replaced with the mean of the grayscale value of the neighborhood pixels. Grayscale normalization maps the grayscale value of the original data to the range of 0-255. The mapping formula is: normalized grayscale value = (original grayscale value - minimum grayscale value) / (maximum grayscale value - minimum grayscale value) × 255.

[0039] The implementation of this technical solution requires a step-by-step approach, ensuring smooth transitions between each step. In the dark current correction stage, the detector needs to continuously acquire data 20 times under X-ray-free conditions, with each acquisition time matching the original data acquisition time. The pixel grayscale mean of the 20 data acquisitions is calculated to obtain the dark current mean matrix. The original data is subtracted pixel by pixel from the dark current mean matrix to eliminate dark current noise generated by the detector's own electron thermal motion. If the original data contains negative grayscale values, they are set to 0. In the bad pixel repair stage, a 3×3 square neighborhood is used. For each pixel, the grayscale mean of its eight neighboring pixels is calculated. If the absolute value of the difference between the pixel's grayscale value and the neighborhood mean exceeds 20% of the neighborhood mean, it is considered a bad pixel, and the grayscale value of the bad pixel is replaced with the neighborhood mean. For pixels located at the image edge, a 2×3 or 3×2 neighborhood is used to ensure sufficient pixels participate in the calculation, avoiding incomplete edge bad pixel repair. In the grayscale normalization stage, the preprocessed original data is traversed first to obtain the maximum and minimum grayscale values ​​of all pixels. If the maximum and minimum values ​​are equal, the grayscale values ​​of all pixels are set to 128; otherwise, the normalized grayscale values ​​are calculated pixel by pixel according to the formula to ensure that the grayscale range of the normalized data is uniform, which facilitates the calculation of the phase correction coefficient and noise suppression weight in the subsequent process, and avoids information extraction deviations caused by differences in the grayscale range of the original data.

[0040] This solution effectively eliminates various noises and distortions through multi-step collaborative preprocessing, providing high-quality data for subsequent information extraction and improving the accuracy of information extraction.

[0041] Traditional technical solutions have the following technical problems: existing information extraction confidence judgment methods do not set thresholds in combination with imaging scenarios, the thresholds are single and fixed, and cannot adapt to the accuracy requirements of different scenarios; when the confidence level is not met, there is no effective adjustment mechanism, which directly abandons or continues extraction, resulting in unstable information quality or low efficiency.

[0042] Based on this, in the step of determining whether the information extraction confidence level meets the preset confidence threshold, the preset confidence threshold is set based on the imaging scenario requirements. The preset confidence threshold for medical dynamic organ imaging scenarios is higher than that for industrial static component detection scenarios. If the information extraction confidence level does not meet the preset confidence threshold, the acquisition frequency of the multi-dimensional interference parameters is adjusted. After increasing the acquisition frequency, the multi-dimensional interference parameters are re-acquired and the phase correction coefficient, noise suppression weight, and information extraction confidence level are calculated until the information extraction confidence level meets the preset confidence threshold.

[0043] This technical solution requires dynamic adjustment of thresholds and acquisition strategies based on scenario characteristics. In the pre-set threshold stage, medical dynamic organ imaging scenarios require the identification of minute airflow channels and demand high information accuracy; the pre-set threshold is set to 0.8. Industrial static component detection scenarios prioritize real-time performance over accuracy; the pre-set threshold is set to 0.7. For special scenarios, the threshold can be set to 0.85. In the confidence judgment and adjustment stage, after the calculation module obtains the information extraction confidence score C, it compares it with the preset threshold. If C ≥ the threshold, subsequent information extraction is performed. If C < the threshold, the acquisition frequency of multi-dimensional interference parameters is first increased by 50%, for example, from 1kHz to 1.5kHz, increasing the data sampling volume to reduce the measurement error of interference parameters. After re-acquiring the interference parameters, the calculation module recalculates P, W, and C, and makes another judgment. If C still does not meet the threshold after three consecutive adjustments, a warning signal is issued, prompting a check of the imaging system, such as whether the grating is offset or the detector is faulty. Simultaneously, the historically optimal P and W values ​​are used to attempt information extraction to avoid complete process interruption.

[0044] This solution ensures the quality of information extraction in different scenarios through scenario-based threshold setting and dynamic adjustment mechanisms, while improving process robustness and avoiding process interruption due to insufficient confidence.

[0045] Traditional technical solutions have the following technical problems: the existing phase correction coefficient calculation is not related to the deformation rate of the imaging object, and is based only on the single parameter of grating phase drift, which cannot compensate for the indirect effect of deformation on phase modulation. The corrected data still has phase deviation, which affects the accuracy of information extraction.

[0046] Based on this, the formula for dynamically calculating the phase correction coefficient is:

[0047] ;

[0048] in, This is the phase correction coefficient, which is dimensionless. This represents the grating phase shift, measured in rad. The deformation rate of the imaged object is expressed in mm / s. The reference deformation rate is the maximum permissible deformation rate adapted to the imaging system. This is the phase compensation adjustment factor, with a value ranging from 0.1 to 0.3; This is the phase drift reference value, in rad, representing the maximum phase fluctuation when the grating is operating stably. It is a natural exponential function with a base of approximately 2.718; This is a cosine function, and the input parameter is in rad.

[0049] The formula design for this technical solution needs to combine physical meaning with engineering practice, and the selection of each parameter must be reasonable and quantifiable. The first part of the formula... As the main correction item, Used to compensate for modulation deviations caused by grating phase drift, when When there is no drift, this item is 1, with no additional correction; when... When it increases, The reduction in magnitude increases the correction force. Used to compensate for the effects of deformation of the imaged object. The speed is set according to the imaging system's performance; for example, 2mm / s for medical scenarios and 5mm / s for industrial scenarios. When the value is 1, this item is 1 and does not affect drift compensation; when When this value increases, the corresponding term decreases, reducing the drift correction strength and avoiding overcorrection. Part Two For fine-tuning items, Take the maximum phase fluctuation when the grating is operating stably (e.g., 0.5 rad). When the value is small (e.g., 0.1 rad), Fine-tuning items provide smaller compensation; when When the value is large (e.g., 1 rad), this term is approximately 0.135, which reduces the compensation of the fine-tuning term and avoids over-fine-tuning. Based on the selected scenario, a value of 0.3 is used for the medical scenario.

[0050] In industrial scenarios, a value of 0.1 is used to balance calibration accuracy and speed. Calculation example: , , , , ,but The correction factor is close to 1, indicating that the interference is small and no significant correction is needed.

[0051] This scheme calculates the phase correction coefficient through multi-parameter linkage, while compensating for the effects of grating drift and object deformation, thereby improving the phase correction accuracy and laying the foundation for subsequent information extraction.

[0052] Traditional technical solutions have the following technical problems: the existing noise suppression weight calculation does not associate with the phase correction coefficient, is based only on the single parameter of noise density, cannot take into account the impact of phase correction on noise distribution, and the weight setting does not match the actual noise suppression requirements, resulting in poor noise filtering effect.

[0053] Based on this, the formula for dynamically calculating the noise suppression weight is:

[0054] ;

[0055] in, The noise suppression weights are dimensionless. Photon statistical noise density, in units of photon / pixel; Maximum permissible noise density, in photons per pixel; The phase correction coefficient as defined in claim 5; This is the noise adjustment factor, with a value ranging from 0.2 to 0.4; The signal-to-noise ratio of the imaging data is calculated from the preprocessed raw data. The calculation method is the ratio of the mean gray level of the signal region to the standard deviation of the gray level of the background region. It is a logarithmic function with base 10.

[0056] The formula design for this technical solution needs to reflect the synergistic effect of noise density, phase correction, and signal-to-noise ratio, and each parameter needs to be accurately calculated. The first part of the formula... As the core weight item, Based on imaging requirements, the settings are as follows: 50 photos per pixel for medical scenarios and 100 photos per pixel for industrial scenarios; when... When it increases, Reduce the weight to avoid information loss due to over-filtering; This reflects the impact of phase correction. The larger the value, the better the phase correction effect; the larger this value, the higher the weight, and the stronger the noise suppression. Part Two This is the signal-to-noise ratio fine-tuning term. During calculation, the signal region is the imaging object region, and the background region is the region without the object, such as... hour, Fine-tuning provides positive compensation; The smaller the value, the smaller the fine-tuning term, thus avoiding excessive weighting under low signal-to-noise ratio. A value of 0.4 is used for medical scenarios and 0.2 for industrial scenarios to adapt to the noise suppression requirements of different scenarios. Calculation example: , , , , ,but The weights are moderate, balancing noise suppression and information preservation.

[0057] This scheme calculates noise suppression weights through multi-parameter linkage, matching the weights with the actual noise state and phase correction effect, thereby improving the targeting and effectiveness of noise filtering.

[0058] Traditional technical solutions have the following technical problems: existing information extraction confidence calculations do not associate phase correction coefficients with noise suppression weights, and are based only on the single parameter of data grayscale mean, which cannot reflect the impact of correction and filtering effects on confidence, resulting in a large deviation between the confidence results and the actual information quality.

[0059] Based on this, the formula for extracting confidence level from dynamically calculated information is:

[0060] ;

[0061] in, The confidence level is extracted for information; it is dimensionless and ranges from 0 to 1. The noise suppression weights as defined in claim 6; The phase correction coefficient as defined in claim 5; The reference phase correction factor is set to 1.0. This is a confidence level adjustment factor, with a value ranging from 0.15 to 0.25; The average grayscale value of the current frame's imaging data; The average grayscale value of the last 10 frames of imaging data; It is a natural exponential function with a base of approximately 2.718.

[0062] The formula design for this technical solution needs to comprehensively consider the effects of correction, filtering, and data stability to ensure that the confidence level accurately reflects the information quality. The first part of the formula... As the core confidence term, The larger the size, the better the noise suppression effect. The larger the value, the better the phase correction effect; the larger this value, the higher the confidence level. hour, This item is This directly reflects the synergistic effect of correction and filtering. Part Two For data stability fine-tuning, It reflects the grayscale difference between the current frame and historical frames. The smaller the difference, the larger this item is, and the higher the confidence level. A value of 0.25 is used for medical scenarios and 0.15 for industrial scenarios to balance stability weights. Calculation example: , , , , ,but The confidence level meets the industrial scenario threshold of 0.7, and information extraction is feasible.

[0063] This scheme calculates confidence levels through multi-factor linkage, accurately reflects information quality, and provides a reliable basis for information extraction decisions.

[0064] Traditional technical solutions have the following technical problems: existing X-ray grating imaging systems have unclear module divisions, data acquisition, interference acquisition and information extraction modules are independent of each other, lack a collaborative working mechanism, data interaction is not smooth, and it is impossible to realize real-time acquisition and dynamic information extraction of multiple interference parameters, resulting in low overall system performance.

[0065] Based on this, this embodiment provides an X-ray grating integrated imaging information extraction system, applied to any of the above-mentioned X-ray grating integrated imaging information extraction methods. It includes a data acquisition module, an interference parameter acquisition module, a calculation module, an information extraction module, and a monitoring and updating module. The data acquisition module includes an X-ray source, a grating assembly, and a detector array. The X-ray source emits an X-ray beam, the grating assembly is positioned between the X-ray source and the imaging object, and the detector array receives X-rays that penetrate the imaging object and the grating assembly, outputting the raw data of the X-ray grating imaging. The interference parameter acquisition module includes a laser interferometer. The system comprises an instrument, a noise analysis unit, and a deformation detection unit. The laser interferometer is used to acquire the phase drift of the grating, the noise analysis unit is used to calculate the photon statistical noise density, and the deformation detection unit is used to obtain the deformation rate of the imaging object. The calculation module is used to dynamically calculate the phase correction coefficient, noise suppression weight, and information extraction confidence level based on multi-dimensional interference parameters. The information extraction module is used to determine whether the information extraction confidence level meets the preset confidence threshold and to perform phase compensation, noise filtering, and information extraction. The monitoring and update module is used to monitor the changes in multi-dimensional interference parameters in real time and, when the update condition is triggered, controls the calculation module to recalculate and update the information extraction process.

[0066] The system modules of this technical solution require clearly defined hardware configurations and functional boundaries, while ensuring real-time and efficient data interaction between modules. In the data acquisition module, the X-ray source is a Varian2100i microfocus X-ray source with an emission energy adjustment range of 20-150kV, set according to the material of the imaging object; the grating assembly consists of a phase grating with a period of 5μm and a thickness of 100μm, and an absorption grating with a period of 5μm and a thickness of 50μm, placed parallel to each other with a spacing of 100mm; the detector array uses a Hamamatsu C12840 flat panel detector with a pixel size of 50μm, a resolution of 2048×2048, and a frame rate of 10fps, connected to the computing module via a CameraLink interface with a data transmission rate of 2.5Gbps. In the interference parameter acquisition module, a ZYGOGPI laser interferometer with a measurement accuracy of 0.1nm and a sampling frequency of 1kHz transmits phase data to the calculation module via a USB 3.0 interface. The noise analysis unit is integrated into the calculation module, calculating noise density through software algorithms. The deformation detection unit uses an FPGA-based image processing module to extract and match feature points in real time, transmitting the deformation rate to the calculation module via a PCIe interface. The calculation module uses a Huawei Kunpeng 920 multi-core processor with integrated DDR5 memory, running a Linux operating system, and synchronously processes the calculations of the three indicators through parallel computing threads. The information extraction module uses GPU-accelerated phase recovery and grayscale inversion algorithms to ensure real-time information extraction. The monitoring and update module controls the parameter monitoring cycle through a timer, triggers updates via an interrupt mechanism, and the modules interact via Ethernet with a latency of less than 10ms.

[0067] This solution, through modular design and collaborative working mechanism, enables real-time acquisition of multiple interference parameters and dynamic information extraction, thereby improving the overall system performance and information extraction quality.

[0068] Traditional technical solutions have the following technical problems: the X-ray source energy in the existing data acquisition module is fixed, which cannot be adapted to imaging objects with different thicknesses and materials; the period of the grating component is mismatched, resulting in poor X-ray modulation effect; the period of the detector array is not matched with that of the grating, so the modulation information cannot be completely acquired, resulting in low data acquisition quality and affecting subsequent information extraction.

[0069] Based on this, the X-ray source in the data acquisition module is a microfocus X-ray source, and the energy of the emitted X-ray beam is adjustable. The adjustment range is set according to the material and thickness of the imaging object. The grating assembly includes a phase grating and an absorption grating. The phase grating and the absorption grating are periodically matched. The phase grating is used to modulate the phase of the X-ray beam, and the absorption grating is used to convert the phase modulation into intensity modulation. The detector array is a flat panel detector. The pixel size of the detector array is adapted to the grating period. The detector array is connected to the computing module through a data transmission interface. The data transmission interface adopts the CameraLink interface, and the transmission rate meets the requirements of real-time imaging.

[0070] The selection and parameter settings of the module hardware in this technical solution must be closely integrated with the imaging requirements. The X-ray source selected is a Varian2100i microfocus X-ray source with a minimum focal size of 5μm, an energy adjustment range of 20-150kV, and a current adjustment range of 0.1-10mA. When imaging the human lung, the energy is set to 40kV and the current to 5mA; when imaging the metal casting, the energy is set to 120kV and the current to 10mA, ensuring that the X-ray beam can penetrate the object and generate effective modulation. In the grating assembly, both the phase grating and the absorption grating are made of silicon, with a period of 5μm. The phase grating is fabricated using photolithography, with a thickness of 100μm, and can generate a phase difference of π / 2. The absorption grating is deposited on a silicon substrate using electroplating, with a thickness of 50μm and a gold layer duty cycle of 50%, ensuring that phase modulation is converted into significant intensity modulation. The two are placed parallel to each other, with a spacing of 100mm set according to the grating period and X-ray energy to avoid modulation superposition interference. The detector array uses a Hamamatsu C12840 flat panel detector with a pixel size of 50μm, which is 10 times the grating period, satisfying the Nyquist sampling theorem to ensure complete acquisition of modulation information. The detector has a frame rate of 10fps, a resolution of 2048×2048, and a single frame data size of approximately 8MB. It is connected to the computing module via a CameraLink interface with a transmission rate of 2.5Gbps, which meets the requirements for real-time data transmission at a frame rate of 10fps without data loss or delay.

[0071] This solution improves the quality of data acquisition through hardware parameter optimization and adaptation, providing high-quality raw data for subsequent information extraction and ensuring the accuracy of information extraction.

[0072] Traditional technical solutions have the following technical problems: existing monitoring and update modules lack clear parameter monitoring cycles and update triggering conditions, parameter monitoring is random, update triggering is blind, and they cannot respond to interference changes in a timely manner, resulting in the information extraction process lagging behind interference changes and a decline in information quality.

[0073] Based on this, the monitoring and update module includes a parameter monitoring unit and an update triggering unit. The parameter monitoring unit collects the current values ​​of multi-dimensional interference parameters at a preset period and calculates the change between the current value and the value of the previous collection period. The update triggering unit pre-stores update conditions. The update condition is that the change of any parameter among the multi-dimensional interference parameters exceeds 15% of the historical average value of the past 10 minutes. If the parameter change meets the update condition, the update triggering unit sends a recalculation instruction to the calculation module and simultaneously controls the information extraction module to pause the current information extraction process. After the calculation module completes the recalculation, the information extraction module restarts the information extraction process based on the new calculation result.

[0074] The implementation of the module functions in this technical solution must ensure real-time monitoring and timely updates. The parameter monitoring unit uses a timer with a preset period of 0.5s. Every 0.5s, it acquires the current values ​​of grating phase drift, photon statistical noise density, and deformation rate of the imaging object from the interference parameter acquisition module. Simultaneously, it reads the parameter values ​​from the previous period and calculates the absolute value of the change (current value - previous period value). The parameter monitoring unit also needs to store parameter data for the past 10 minutes and calculate the historical average value using a sliding window algorithm. The window size is 1200 (10 minutes / 0.5s). Every time the parameter data is updated, the historical average value is updated synchronously. The update trigger unit pre-stores update conditions and compares the change in each parameter with 15% of the historical average. If the change in any parameter exceeds this threshold, the update condition is deemed met. The update trigger unit sends a recalculation command, containing the current parameter value, to the calculation module via Ethernet. Simultaneously, it sends a pause command to the information extraction module, which pauses information extraction for the current frame and saves the processed data. Upon receiving the command, the calculation module recalculates the phase correction coefficient, noise suppression weight, and information extraction confidence level within 50ms and sends the results to the information extraction module. After receiving the new results, the information extraction module restarts the information extraction process and processes the current frame data based on the new results. If the parameter change triggers an update three times consecutively, the update trigger unit sends a warning signal to the system, prompting the operator to check for interference sources.

[0075] This solution ensures that the information extraction process is always adapted to the current interference state through precise monitoring and timely updates, thus guaranteeing stable information quality.

[0076] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method of X-ray grating-based imaging information extraction, characterized in that, The method comprises the following steps: acquiring raw data of X-ray grating imaging and pre-processing the raw data; collecting multi-dimensional interference parameters, including grating phase shift, photon statistical noise density and imaging object deformation rate; based on the multi-dimensional interference parameters, dynamically calculating phase correction coefficient, noise suppression weight and information extraction confidence; judging whether the information extraction confidence meets the pre-set confidence threshold, if yes, performing phase compensation on the pre-processed raw data based on the phase correction coefficient, and then performing noise filtering in combination with the noise suppression weight, and finally extracting the structural information and density information of the imaging object; real-time monitoring the change state of the multi-dimensional interference parameters, if the change amount of any parameter in the multi-dimensional interference parameters triggers the update condition, then recalculating the phase correction coefficient, noise suppression weight and information extraction confidence, and updating the information extraction process according to the new calculation result.

2. The method of claim 1, wherein, In the step of collecting multi-dimensional interference parameters, the grating phase shift is collected by a laser interferometer, the sampling frequency of the laser interferometer matches the period of the X-ray grating, and the change amount of the grating phase with time is obtained after Fourier transform of the collected data; the photon statistical noise density is obtained by calculating the pixel gray variance of continuous multiple frames of pre-processed raw data, and the background area in the imaging area without imaging object is selected as the sample area during calculation; the imaging object deformation rate is obtained by a feature point matching algorithm, the feature points are extracted in the current frame of imaging data, then the corresponding feature points in the previous frame of imaging data are matched, the ratio of the feature point displacement amount to the interval time between the two frames is calculated, and the average value of the ratios of all feature points is taken as the imaging object deformation rate.

3. The method of claim 1, wherein, The step of pre-processing the raw data includes dark current correction, bad pixel repair and gray scale normalization, the dark current correction is realized by subtracting the average value of the dark current data output by the detector when there is no X-ray radiation from the raw data; the bad pixel repair adopts a neighborhood interpolation method, and the gray value of a pixel deviating from the average value of the neighborhood by more than a preset deviation value is replaced by the average gray value of the neighborhood pixels; the gray scale normalization maps the gray value of the raw data to the range of 0-255.

4. The method of claim 1, wherein, In the step of judging whether the information extraction confidence meets the pre-set confidence threshold, the pre-set confidence threshold is set based on the imaging scene requirement, and the pre-set confidence threshold in the medical dynamic organ imaging scene is higher than that in the industrial static part detection scene; if the information extraction confidence does not meet the pre-set confidence threshold, the collection frequency of the multi-dimensional interference parameters is adjusted, the multi-dimensional interference parameters are re-collected after increasing the collection frequency, and the phase correction coefficient, the noise suppression weight and the information extraction confidence are calculated, until the information extraction confidence meets the pre-set confidence threshold.

5. The method of claim 1, wherein, The formula for dynamically calculating the phase correction coefficient is: ； wherein, is a phase correction coefficient, dimensionless; is a grating phase drift, in rad; is an imaging object deformation rate, in mm / s; is a reference deformation rate, is a maximum allowed deformation rate adapted by the imaging system; is a phase compensation adjustment factor, with a value range of 0.1 to 0.3; is a phase drift reference value, in rad, is a maximum phase fluctuation amount when the grating is stably working; is a natural exponential function, with a base of about 2.718; is a cosine function, with an input parameter in rad.

6. The method of claim 1, wherein, The formula for dynamically calculating the noise suppression weight is: ； wherein, is a noise suppression weight, dimensionless; is a photon statistical noise density, unit: photon / pixel; is a maximum allowed noise density, unit: photon / pixel; is a phase correction coefficient defined in claim 5; is a noise adjustment factor, with a value range of 0.2 to 0.4; is a signal-to-noise ratio of the imaging data, calculated from the pre-processed original data, with a calculation method of a ratio of a mean gray value of a signal region to a standard deviation of a gray value of a background region; is a logarithm function with a base of 10.

7. The method of claim 1, wherein, The formula for dynamically calculating the information extraction confidence is: ； wherein, is the information extraction confidence, dimensionless, with a value range of 0 to 1; is the noise suppression weight defined in claim 6; is the phase correction coefficient defined in claim 5; is the reference phase correction coefficient, with a value of 1.0; is the confidence adjustment factor, with a value range of 0.15 to 0.25; is the gray level mean of the current frame imaging data; is the gray level mean of the last 10 frames of imaging data; is the natural exponential function, with a base of about 2.

718.

8. An X-ray grating comprehensive imaging information extraction system applied to an X-ray grating comprehensive imaging information extraction method according to any one of claims 1 to 7, comprising a data acquisition module, an interference parameter acquisition module, a calculation module, an information extraction module and a monitoring and updating module, the data acquisition module comprising an X-ray source, a grating assembly and a detector array, the X-ray source being used for emitting an X-ray beam, the grating assembly being arranged between the X-ray source and an imaging object, the detector array being used for receiving X-rays penetrating through the imaging object and the grating assembly and outputting original data of X-ray grating imaging; the interference parameter acquisition module comprising a laser interferometer, a noise analysis unit and a deformation detection unit, the laser interferometer being used for acquiring a grating phase drift amount, the noise analysis unit being used for calculating a photon statistical noise density, and the deformation detection unit being used for acquiring an imaging object deformation rate; the calculation module being used for dynamically calculating a phase correction coefficient, a noise suppression weight and an information extraction confidence level based on multi-dimensional interference parameters; the information extraction module being used for judging whether the information extraction confidence level meets a preset threshold value, and performing phase compensation, noise filtering and information extraction; and the monitoring and updating module being used for monitoring a multi-dimensional interference parameter change state in real time, and controlling the calculation module to recalculate and update an information extraction process when an update condition is triggered.

9. An X-ray grating-based imaging information extraction system according to claim 8, characterized in that, The X-ray source in the data acquisition module adopts a micro-focus X-ray source, the emitted X-ray beam energy is adjustable, and the adjustment range is set according to the material and thickness of the imaging object; the grating assembly comprises a phase grating and an absorption grating, the periods of the phase grating and the absorption grating are matched, the phase grating is used for modulating the phase of the X-ray beam, and the absorption grating is used for converting the phase modulation into intensity modulation; the detector array adopts a flat panel detector, the pixel size of the detector array is adapted to the grating period, and the detector array is connected to the calculation module through a data transmission interface, the data transmission interface adopts a CameraLink interface, and the transmission rate meets the real-time imaging requirement.

10. The X-ray grating synthetic imaging information extraction system of claim 8, wherein, The monitoring and updating module comprises a parameter monitoring unit and an update triggering unit, the parameter monitoring unit acquires current values of the multi-dimensional interference parameters at a preset period, and calculates the change amount of the current values and the values of the last acquisition period; The update triggering unit pre-stores an update condition, the update condition being that the change amount of any parameter in the multi-dimensional interference parameters exceeds 15% of the near-10-minute historical average value, if the parameter change amount meets the update condition, the update triggering unit sends a recalculation instruction to the calculation module, and controls the information extraction module to pause the current information extraction process, after the calculation module completes the recalculation, the information extraction module restarts the information extraction process according to the new calculation result.

Images