Dual-energy x-ray bone densitometry method based on flat panel detector

Abstract

The application discloses a dual-energy ray bone density evaluation method based on a flat panel detector and relates to the technical field of ray evaluation.The direct conversion mode of the flat panel detector is used to realize high-precision collection of low-energy and high-energy ray images, and noise and resolution loss caused by traditional indirect imaging are significantly reduced;one-time and two-time corrections are completed in an edge computing node, the problem of gray scale discontinuity and energy ratio distortion in the joint area of bones and soft tissues is effectively solved, and the authenticity of the image boundary is improved;further dynamic smoothing processing of the energy ratio is combined to obtain a stable ratio image, the stability and repeatability of bone density measurement are significantly enhanced, and more accurate bone density evaluation is realized.

Description

Technical Field

[0001] This invention relates to the field of radiographic assessment technology, and in particular to a dual-energy X-ray bone mineral density assessment method based on a flat panel detector. Background Technology

[0002] With the accelerating aging of the population, the incidence of osteoporosis and related bone diseases is increasing year by year. Bone mineral density measurement, as an important indicator for assessing bone health and predicting fracture risk, has received widespread attention in the field of medical imaging. Traditional methods for assessing bone mineral density mainly include ultrasound measurement, single-energy X-ray absorptiometry (SDA), and dual-energy X-ray absorptiometry (DAX). Among them, DAX has become the mainstream clinical application because it can distinguish between bone tissue and soft tissue and quantitatively analyze bone mineral density.

[0003] However, existing dual-energy X-ray absorptiometry (DAX) systems are typically based on scintillator detectors and indirect conversion methods, which suffer from signal attenuation, insufficient spatial resolution, and low contrast during image acquisition, especially in the bone-soft tissue interface region where errors are prone to occur. Furthermore, some systems rely on centralized computing platforms for post-processing, making it difficult to handle the real-time computational demands of large-scale data, thus limiting their application in high-throughput screening and rapid diagnosis scenarios in clinical settings. Summary of the Invention

[0004] In view of the problems existing in the dual-energy X-ray bone mineral density measurement technology, this invention is proposed.

[0005] Therefore, the problem to be solved by this invention is how to improve the accuracy and robustness of bone mineral density calculation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a dual-energy X-ray bone mineral density assessment method based on a flat panel detector, comprising: acquiring low-energy X-ray image datasets and high-energy X-ray image datasets respectively using a direct conversion method of a flat panel detector, and performing a first correction at an edge computing node; establishing an attenuation inversion function in the edge computing node based on the first-corrected low-energy X-ray image dataset and high-energy X-ray image dataset, and performing a second correction on the boundary region between bone and soft tissue; calculating an energy ratio sequence based on the second-corrected low-energy X-ray image dataset and high-energy X-ray image dataset, and performing dynamic smoothing processing on the energy ratio in the edge computing node to obtain a stable ratio image; and separating bone tissue images and soft tissue images based on the stable ratio image, and calculating bone mineral density values ​​based on the separated bone tissue images.

[0008] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the step of performing a correction at the edge computing node includes: acquiring multiple frames of low-energy X-ray images and high-energy X-ray images, adding a time index to each frame; for each pixel, statistically analyzing the difference sequence between the grayscale value and the mean grayscale value of the same pixel's neighborhood in consecutive frames; if the absolute difference in the difference sequence exceeds the standard deviation of the neighborhood difference for more than one-third of the frames, it is marked as a defective pixel; when the proportion of defective pixels appearing in the total frames exceeds a preset proportion, it is determined to be a high-level defective pixel, otherwise it is a low-level defective pixel; for high-level defective pixels, neighborhood block fitting is used for repair, and the original pixel is replaced; for low-level defective pixels, grayscale values ​​are generated by interpolation along the row and column directions, and the original pixel is replaced; the repaired pixel values ​​are directly used to correct the low-energy X-ray image dataset and the high-energy X-ray image dataset.

[0009] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the step of establishing an attenuation inversion function in the edge computing node includes: calculating the horizontal and vertical neighborhood gradients for each pixel in the corrected low-energy and high-energy images, identifying candidate abnormal pixels at the boundaries; and calculating the energy ratio of each pixel in the corrected low-energy X-ray image and the high-energy X-ray image. ; Calculate the average energy ratio of each pixel within a defined neighborhood. Based on the local gradient direction information, the energy ratio difference and the local gradient direction are combined to form a pixel-level attenuation inversion function:

[0010] ;

[0011] ;

[0012] in, These are pixel-level attenuation inversion function values; The difference between the ratios; The direction of the pixel gradient; These are the direction-weighted coefficients.

[0013] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the secondary correction of the boundary region between bone and soft tissue includes: adjusting the boundary candidate abnormal pixels based on the pixel-level attenuation inversion function value so that the difference between the energy ratio of the boundary candidate abnormal pixels and the neighborhood reference ratio is less than a preset difference threshold; performing a weighted average correction along the neighborhood gradient direction in the principal gradient direction of each boundary candidate abnormal pixel, maintaining the gray level continuity of the bone and soft tissue boundary; if, among the continuous boundary candidate abnormal pixels, there is a boundary candidate abnormal pixel whose energy ratio change exceeds the neighborhood standard deviation after weighted average correction, then the gray level is re-corrected to the mean gray level of the neighborhood pixels to obtain the low-energy X-ray image dataset and the high-energy X-ray image dataset after secondary correction.

[0014] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the determination of the boundary candidate abnormal pixels includes: in the edge computing node, comparing the changes in the horizontal and vertical neighboring pixel values ​​of the low-energy X-ray image and the high-energy X-ray image pixel by pixel; when the change direction of a pixel is opposite to that of the pixel value of its horizontal or vertical neighboring neighbor and the magnitude exceeds a preset change ratio of the average change of the neighboring neighboring neighbor, it is marked as a boundary candidate abnormal pixel.

[0015] As a preferred embodiment of the dual-energy X-ray bone mineral density assessment method based on a flat panel detector described in this invention, the step of correcting the grayscale value to the mean grayscale value of neighboring pixels is performed using the following calculation formula:

[0016] ;

[0017] in, The pixel gray value is the result of correction to the mean gray value of neighboring pixels; The average grayscale value of the neighboring pixels; The pixel grayscale value is corrected by weighted average. For re-correction factors; It is a set of local neighborhood pixels.

[0018] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the dynamic smoothing of energy ratios in the edge computing nodes includes: if an energy ratio in the energy ratio sequence is inconsistent with the direction of change of neighboring pixels in the same row or column, it is marked as a candidate abnormal pixel; combined with the energy ratio changes of the neighborhood in the same direction, the energy ratio of the candidate abnormal pixel is adjusted to the extended value of the neighborhood trend; the consistency of the trend of each candidate abnormal pixel in the horizontal and vertical directions is checked, and the candidate abnormal pixels whose trend direction deviates from the neighborhood pattern are fine-tuned to ensure the stability of the grayscale structure of the bone boundary and soft tissue transition area, and a stable ratio image for each frame is output.

[0019] As a preferred embodiment of the dual-energy X-ray bone density assessment method based on a flat panel detector described in this invention, the separation of bone tissue image and soft tissue image includes: for each pixel in the stable ratio image, calculating the difference between the energy ratio and the median difference between the energy ratio and the median difference of the neighboring pixel ratios in the horizontal and vertical directions; when the energy ratio is higher than the median difference of the neighboring pixel ratios by more than a preset difference threshold, it is determined to be a bone tissue pixel; the remaining pixels are determined to be soft tissue pixels; for pixels in the boundary area between bone tissue pixels and soft tissue pixels, judging the consistency between the energy ratio and the neighborhood direction of the bone tissue pixel; if they are consistent, they are classified as bone tissue, otherwise they are classified as soft tissue, ensuring the continuity of the bone tissue boundary; based on the pixel determination results, bone tissue pixels are used to form a bone tissue image, and soft tissue pixels are used to form a soft tissue image.

[0020] As a preferred embodiment of the dual-energy X-ray bone mineral density assessment method based on a flat panel detector described in this invention, the step of calculating bone mineral density values ​​based on the separated bone tissue images includes: extracting the energy ratio of each pixel in the bone tissue image from the stable ratio image, and statistically analyzing the differences in energy ratios with neighboring pixels in the horizontal and vertical directions to form a local bone resorption index; combining the local bone resorption index with a pre-calibrated relationship between low-energy and high-energy ratios and bone mineral density, mapping the local index of each pixel to a specific bone mineral density value to form a pixel-level bone mineral density value; and combining the bone mineral density values ​​of all bone tissue pixels to generate a bone mineral density map, while setting the grayscale of soft tissue pixels to zero.

[0021] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program instructions, when executed by the processor, implement the steps of the dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in the first aspect of the present invention.

[0022] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in the first aspect of the present invention.

[0023] The beneficial effects of this invention are as follows: This invention achieves high-precision acquisition of low-energy and high-energy X-ray images through direct conversion of flat panel detectors, significantly reducing noise and resolution loss caused by traditional indirect imaging; and completes primary and secondary corrections in edge computing nodes, effectively solving the problems of grayscale discontinuity and energy ratio distortion in the bone-soft tissue interface area, thereby improving the realism of image boundaries; further, combined with dynamic smoothing processing of energy ratios, a stable ratio image is obtained, significantly enhancing the stability and repeatability of bone density measurement, and achieving more accurate bone density assessment.

[0024] This invention not only improves the accuracy and robustness of the test results, but also takes into account real-time performance and scalability, and can be widely used in osteoporosis screening and clinical diagnosis, thus having important medical application value. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a dual-energy X-ray bone mineral density assessment method based on a flat panel detector. Detailed Implementation

[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0030] Figure 1 This is a flowchart of a dual-energy X-ray bone mineral density assessment method based on a flat panel detector according to an embodiment of the present invention. Figure 1 As shown, the dual-energy X-ray bone mineral density assessment method based on a flat panel detector includes:

[0031] S1: Using the direct conversion method of the flat panel detector, low-energy ray image datasets and high-energy ray image datasets are acquired respectively, and a correction is performed at the edge computing node.

[0032] S1.1: Acquire multiple frames of low-energy ray images and high-energy ray images, and add a time index to each frame.

[0033] S1.2: For each pixel, the difference sequence between the gray value in consecutive frames and the gray value mean of the same pixel neighborhood (e.g., 3×3 pixels) is calculated. This not only considers the temporal variation of a single pixel, but also introduces the stability of the spatial neighborhood as a reference benchmark, avoiding deviations caused by isolated pixel anomalies.

[0034] If the absolute difference of more than one-third of the frames in the difference sequence is greater than the standard deviation of the neighborhood difference, it is marked as a defective pixel; when the proportion of defective pixels appearing in the total number of frames exceeds a preset proportion (e.g., 0.4), it is determined to be a high-level defective pixel, otherwise it is a low-level defective pixel.

[0035] It should be noted that conventional pixel stability assessment methods often rely solely on grayscale threshold comparisons within a single frame of the image. This approach can lead to numerous false positives when noise levels are high or radiation doses are insufficient. The method of this invention, by simultaneously analyzing both temporal and spatial dimensions, ensures the accuracy and robustness of defective pixel labeling.

[0036] S1.3: For high-level defect pixels, neighborhood block fitting is used for repair and the original pixel is replaced; for low-level defect pixels, grayscale values ​​are generated by interpolation along the row and column directions and the original pixel is replaced; the repaired pixel values ​​are directly updated in the low-energy ray image dataset and the high-energy ray image dataset.

[0037] It should be noted that after completing the defect pixel labeling and classification, different repair strategies need to be adopted for defect pixels of different levels to ensure the accuracy and integrity of the low-energy X-ray image dataset and the high-energy X-ray image dataset. Conventional repair methods often use a uniform interpolation method, such as bilinear interpolation or nearest neighbor interpolation. This method is prone to producing obvious artifacts in high-level defect areas, thus affecting the subsequent calculation of bone mineral density. This invention uses two types of repair methods by differentiating levels, thereby improving the repair effect. Specifically:

[0038] Specifically, a large neighborhood block, such as a 5×5 or 7×7 pixel range, is selected as the center of the high-level defect pixel. The grayscale values ​​of normal pixels are extracted from this neighborhood block, and the local grayscale distribution is determined using polynomial fitting or least squares fitting. The reasonable grayscale value of the defect pixel is then calculated and replaced with the original defect pixel value. Neighborhood block fitting fully utilizes the local grayscale distribution characteristics, ensuring that the repair result is consistent with the grayscale change trend of the surrounding tissue, and avoiding abrupt transitions caused by simple interpolation.

[0039] For low-level defect pixels, two adjacent normal pixels are selected to the left or right, or above or below, the defect pixel in both the horizontal and vertical directions. The grayscale value of the defect pixel is calculated using linear interpolation, and the interpolation result is used as the replacement value to update the image. Since low-level defect pixels are isolated anomalies in space and time, and their number and distribution range are small, simple directional interpolation can ensure repair efficiency while preserving the detailed features of the original image.

[0040] After the repair is complete, the new pixel values ​​obtained from the replacement are directly updated in both the low-energy X-ray image dataset and the high-energy X-ray image dataset. Through this update method, the repair process is no longer merely a post-processing step, but rather an integral part of the image dataset construction, ensuring that subsequent inversion function calculations and bone tissue image separation are performed on high-quality data.

[0041] S2: Based on the low-energy ray image dataset and high-energy ray image dataset after primary correction, an attenuation inversion function is established in the edge computing node to perform secondary correction on the boundary region between bone and soft tissue.

[0042] S2.1: Establish a decay inversion function in the edge computing nodes.

[0043] It should be noted that in the low-energy ray image dataset and the high-energy ray image dataset after one correction, the response of each pixel under different energy conditions is not completely consistent. Since the ray absorption of bone and soft tissue differs significantly under dual-energy conditions, it is necessary to construct an attenuation inversion function to achieve pixel-level differential mapping. Traditional methods typically approximate the distinction between bone and soft tissue using only the energy ratio, but this approach ignores local gradient features and is prone to blurring or error accumulation in boundary regions. This invention proposes to consider both the ratio difference and gradient direction information in the edge computing nodes to establish an attenuation inversion function, thereby obtaining attenuation mapping values ​​at the pixel level that are more consistent with physical laws, providing a precise quantitative basis for subsequent boundary correction. Specifically:

[0044] First, in the edge computing nodes, the gray-level difference is calculated pixel-by-pixel for the low-energy and high-energy ray images, and the direction of change of pixel values ​​in the horizontal and vertical neighborhoods is compared. When the direction of gray-level change of a pixel is opposite to the direction of change of most pixel values ​​in its horizontal or vertical neighborhood and the magnitude exceeds a preset change ratio of the neighborhood average, it is marked as a candidate abnormal pixel for the boundary. This judgment strategy can effectively identify the location of gray-level abrupt changes, avoid the boundary errors caused by filtering solely through global thresholds, and thus improve the accuracy of boundary recognition.

[0045] Secondly, the energy ratio of each pixel in the low-energy ray image and the high-energy ray image after one correction is calculated. The energy ratio is defined as the ratio of the grayscale response of the same pixel under low-energy and high-energy ray conditions. By calculating the energy ratio, the differences between bone and soft tissue under dual-energy conditions can be amplified, thus providing a mathematical basis for differentiation. However, the energy ratio alone is sensitive to noise, so it is necessary to further calculate the mean energy ratio within a defined local neighborhood. By calculating the neighborhood mean, a locally stable reference ratio can be obtained for subsequent difference comparisons.

[0046] Furthermore, the average energy ratio of each pixel is calculated within a defined neighborhood. Based on the local gradient direction information, the energy ratio difference and the local gradient direction are combined to form a pixel-level attenuation inversion function:

[0047] ;

[0048] ;

[0049] in, This is the pixel-level attenuation inversion function value, representing the attenuation mapping reference value of this pixel under low-energy and high-energy conditions; The ratio difference indicates the degree to which the pixel deviates from the local mean; The pixel gradient direction represents the angle between the direction of grayscale change and the horizontal direction. These are the direction-weighted coefficients.

[0050] As can be seen, the pixel-level attenuation inversion function provides direction sensitivity in boundary regions with significant grayscale changes, thus ensuring that grayscale transitions in different directions better conform to actual anatomical structures. The establishment of the attenuation inversion function not only provides reference values ​​for pixel grayscale attenuation mapping under low-energy and high-energy conditions, but also provides accurate function inputs for subsequent secondary correction. Compared with traditional methods relying on a single ratio or global threshold, this method maintains computational stability and accuracy even under extreme conditions (such as extremely narrow regions at the boundary between cortical bone and soft tissue).

[0051] S2.2: Perform secondary correction on the junction area between bone and soft tissue.

[0052] Gray-level differences between bones and soft tissues often manifest as abrupt changes at image boundaries. A single correction only ensures overall pixel consistency and cannot completely eliminate local deviations at the boundaries. Without secondary correction, artifacts or gray-level discontinuities can easily occur at the boundaries, directly affecting the accurate calculation of bone mineral density. Therefore, the purpose of secondary correction is to adjust the energy ratio of candidate anomalous pixels at the boundary to more closely approximate the neighborhood reference value, while preserving the details of the bone tissue microstructure.

[0053] S2.2.1: Adjust the boundary candidate anomalous pixels based on the pixel-level attenuation inversion function value, so that the difference between the energy ratio of the boundary candidate anomalous pixels and the neighborhood reference ratio is less than a preset difference threshold. This method ensures that the boundary candidate anomalous pixels can numerically approach the neighborhood mean, avoiding image discontinuities caused by excessive deviation.

[0054] S2.2.2: In the main gradient direction of each candidate abnormal pixel at the boundary, a weighted average correction is performed with the gray level of the neighboring pixels along the gradient direction (i.e., the gray level is fused with the current pixel gray level by weighted averaging) to maintain the gray level continuity of the bone and soft tissue boundary, while preserving the details of the bone microstructure as much as possible.

[0055] This gradient-direction correction method can maintain the grayscale continuity of the bone and soft tissue boundaries, while avoiding structural blurring caused by lateral interpolation. Especially in complex microstructure regions such as trabeculae, gradient-direction smoothing can preserve detailed information and prevent the loss of tissue boundaries due to over-correction.

[0056] S2.2.3: If, among the continuous boundary candidate anomalous pixels, there are boundary candidate anomalous pixels whose energy ratio changes beyond the neighborhood standard deviation after weighted average correction, then the gray level is re-corrected to the gray level mean of the neighborhood pixels to obtain the low-energy ray image dataset and high-energy ray image dataset after secondary correction.

[0057] The grayscale value is corrected to the mean grayscale value of neighboring pixels using the following calculation formula:

[0058] ;

[0059] in, The pixel gray value is the result of correction to the mean gray value of neighboring pixels; The average grayscale value of the neighboring pixels; The pixel grayscale value is corrected by weighted average. For re-correction factors; It is a set of local neighborhood pixels.

[0060] This method effectively prevents strong local noise or abnormal pixels from disrupting the overall continuity of the boundary. After the above processing, a low-energy X-ray image dataset and a high-energy X-ray image dataset with secondary correction are obtained, laying a high-quality data foundation for subsequent energy ratio calculation and bone tissue separation.

[0061] S3: Calculate the energy ratio sequence based on the low-energy ray image dataset and high-energy ray image dataset after secondary correction, and perform dynamic smoothing on the energy ratio in the edge computing node to obtain a stable ratio image.

[0062] S3.1: Calculate the energy ratio sequence.

[0063] For each pixel location, a preliminary energy ratio map is generated based on the high-energy and low-energy gray values ​​after secondary correction in S2.2, reflecting the difference in absorption between bone and soft tissue. Compared with traditional single-energy images, the preliminary energy ratio map can highlight the attenuation difference between bone and soft tissue regions in the same frame of data, reducing the computational complexity of subsequent separation.

[0064] According to the acquisition sequence, each frame of the ratio image is numbered to form a ratio time series. In the ratio time series, each pixel corresponds to a ratio change trajectory, reflecting the ratio stability throughout the scanning process.

[0065] In the ratio sequence of each pixel, the direction of change of the pixel in the time series is analyzed and compared with the direction of change of neighboring pixels in the same row and column. If the direction of change of the ratio of a pixel at multiple time points is inconsistent with the direction of change of neighboring pixels in the same row or column, the pixel is marked as a candidate anomalous pixel. Such pixels usually correspond to local noise, micro-vibrations during the scanning process, or transient response anomalies of the detector. If left untreated, artifacts or tomography can easily be generated in the ratio image. Through this marking mechanism, anomalies can be identified simultaneously in both spatial and temporal dimensions, avoiding missed detections or false positives caused by judging based on only a single dimension.

[0066] S3.2: Perform dynamic smoothing of energy ratios in edge computing nodes.

[0067] After obtaining the energy ratio sequence, the candidate anomalous pixels must be corrected to eliminate the interference of local mutations on the overall structural stability.

[0068] First, by combining the energy ratio changes of neighboring regions in the same direction, if the ratio change of candidate anomalous pixels significantly deviates from the neighborhood trend, the energy ratio of the candidate anomalous pixels is adjusted to an extended value of the neighborhood trend. This method avoids the interference of single-point mutations on the overall sequence while ensuring that the corrected values ​​have temporal continuity and spatial consistency. Unlike conventional mean-based alternative methods, this trend extension correction can more realistically reflect the physical changes of local tissues in the time dimension.

[0069] Secondly, the consistency of the trend of each candidate anomalous pixel in the horizontal and vertical directions is checked, and the candidate anomalous pixels whose trend direction deviates from the neighborhood pattern are fine-tuned to ensure the stability of the gray-scale structure of the bone boundary and soft tissue transition area, and the stable ratio image of each frame is output.

[0070] Specifically, the method involves calculating the ratio change pattern of the pixel in both the horizontal and vertical neighborhoods and comparing it with the change pattern of candidate anomalous pixels. If the ratio change of a candidate anomalous pixel deviates significantly from the neighborhood pattern in a certain direction, the comparison value in that direction is corrected again to maintain consistency with the neighborhood pattern. This consistency check ensures that the ratio image maintains structural stability in both the horizontal and vertical directions, avoiding grayscale breaks or artifacts caused by unidirectional correction.

[0071] In the final stage of dynamic smoothing, the ratio results after trend extension correction and consistency correction are output as a stable ratio image.

[0072] Stable ratio images possess not only spatial boundary continuity but also temporal sequence consistency, thus providing reliable data input for subsequent bone and soft tissue separation. Particularly in complex regions such as trabeculae and the cortical-bone junction, stable ratio images significantly reduce artifact interference, ensuring the accuracy and stability of bone mineral density calculations. Compared to traditional methods, this method exhibits significant advantages in both noise robustness and detail fidelity.

[0073] S4: Based on the stable ratio image, separate the bone tissue image and the soft tissue image, and calculate the bone mineral density value based on the separated bone tissue image.

[0074] S4.1: Separate bone tissue images and soft tissue images.

[0075] S4.1.1: For each pixel in the stable ratio image, calculate the difference between the energy ratio and the median difference between the energy ratio and the ratios of neighboring pixels in the horizontal and vertical directions. When the energy ratio is higher than the median difference of the neighboring pixel ratio by more than a preset difference threshold, it is determined to be a bone tissue pixel; the remaining pixels are determined to be soft tissue pixels. The use of the median can avoid interference from extreme values ​​and make the determination criteria more robust.

[0076] The preset difference threshold is derived from the experimental calibration of the differences in attenuation characteristics between bone tissue and soft tissue at different energy levels, which can ensure that the distinction between bone tissue and soft tissue has a clear numerical boundary, rather than relying on fuzzy judgment.

[0077] This method allows bone tissue pixels to be identified and marked point by point in the image, avoiding the mistaken identification of high-noise pixels as bone tissue.

[0078] When processing the remaining pixels that do not meet the threshold condition, these pixels are classified as soft tissue pixels. Since soft tissue shows relatively small absorption differences under low-energy and high-energy rays, the energy ratio is usually not significantly higher than the median of the neighborhood. Therefore, this criterion can accurately identify soft tissue pixels. Conventional bone density analysis often relies on overall grayscale interval segmentation, while this method uses pixel-by-pixel median difference comparison, which improves the stability of boundary recognition. This determination strategy avoids over-reliance on noise from single-frame images and preserves detailed information during the differentiation of bone and soft tissue, improving the accuracy of subsequent bone mineral density calculations.

[0079] S4.1.2: For pixels in the boundary region between bone tissue pixels and soft tissue pixels, determine whether the energy ratio is consistent with the direction of the neighboring bone tissue pixels, such as continuously increasing in the horizontal direction or continuously decreasing in the vertical direction. If they are consistent, they are classified as bone tissue. If the trend of energy ratio change is inconsistent with the direction of the neighboring bone tissue pixels, the pixel is classified as a soft tissue pixel, thereby preventing the unreasonable expansion of the bone tissue boundary and ensuring the continuity of the bone tissue boundary.

[0080] S4.1.3: Based on the pixel determination results, bone tissue pixels are used to form bone tissue images, and soft tissue pixels are used to form soft tissue images.

[0081] Bone tissue images reflect the distribution of skeletal structures within a stable ratio image, while soft tissue images preserve the distribution information of tissues other than bone. This pixel-level separation not only provides clean bone tissue image input for subsequent bone mineral density calculations but also allows for separate analysis of soft tissues when needed.

[0082] S4.2: Calculate bone mineral density values ​​based on the separated bone tissue images.

[0083] S4.2.1: For each pixel in the bone tissue image, extract the energy ratio in the stable ratio image, and statistically analyze the difference in energy ratio with neighboring pixels in the horizontal and vertical directions to form a local bone resorption index.

[0084] The energy ratio difference can be calculated by taking each pixel in the bone tissue image, calculating the absolute difference between the energy ratio of the pixel and the energy ratios of its horizontal and vertical neighboring pixels, and using the average of the differences as the statistical result of the pixel's energy ratio difference.

[0085] Local bone resorption indices reflect the differences in X-ray attenuation characteristics between a pixel and its surrounding bone tissue. A higher value indicates a significant difference in X-ray absorption characteristics between the pixel and its neighborhood, potentially suggesting localized bone density abnormalities. Conversely, a lower value indicates that the pixel's absorption characteristics are similar to its neighborhood, indicating a more homogeneous bone tissue structure. This method allows for the construction of local bone tissue absorption features at the pixel level, providing refined parameters for calculating bone mineral density.

[0086] S4.2.2: By combining the local bone resorption index with the pre-calibrated low-energy to high-energy ratio and the relationship with bone mineral density, the local index of each pixel is mapped to a specific bone mineral density value, forming a pixel-level bone mineral density value.

[0087] Specifically, the mapping process relies on a pre-calibrated correspondence between the low-energy ray ratio, the high-energy ray ratio, and bone mineral density. This relationship is obtained through standard samples or clinical calibration experiments; that is, under different known bone mineral densities, the corresponding low-energy and high-energy ratios are measured to establish an empirical curve or function. Then, by substituting the local bone resorption index of each pixel into the above relationship model, pixel-level bone mineral density values ​​can be obtained.

[0088] As can be seen, this invention avoids dependence on the average value of the entire region, and instead achieves high-resolution bone density distribution calculation through pixel-by-pixel mapping. This method can reflect the differences in mineral content within the fine structure of bones, providing support for the early detection of diseases such as osteoporosis.

[0089] S4.2.3: Combine the bone mineral density values ​​of all bone tissue pixels to generate a bone mineral density map. Set the grayscale of soft tissue pixels to zero to avoid interference with bone density distribution. The final bone mineral density map visually displays the density changes within the bone region while eliminating the influence of soft tissue regions, making the results more accurate.

[0090] This embodiment also provides a computer device applicable to the dual-energy X-ray bone mineral density assessment method based on a flat panel detector, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the dual-energy X-ray bone mineral density assessment method based on a flat panel detector as proposed in the above embodiment.

[0091] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0092] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the dual-energy X-ray bone mineral density assessment method based on a flat panel detector as proposed in the above embodiments.

[0093] In summary, this invention achieves high-precision acquisition of low-energy and high-energy X-ray images through direct conversion using a flat panel detector, significantly reducing noise and resolution loss associated with traditional indirect imaging. Furthermore, by performing primary and secondary corrections at edge computing nodes, it effectively solves the problems of grayscale discontinuity and energy ratio distortion in the bone-soft tissue interface region, thereby improving the realism of image boundaries. Further, by combining dynamic smoothing of the energy ratio, a stable ratio image is obtained, significantly enhancing the stability and repeatability of bone density measurement and achieving more accurate bone density assessment.

[0094] This invention not only improves the accuracy and robustness of the test results, but also takes into account real-time performance and scalability, and can be widely used in osteoporosis screening and clinical diagnosis, thus having important medical application value.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dual-energy X-ray bone mineral density assessment method based on a flat panel detector, characterized in that: include: Using a direct conversion method with a flat panel detector, low-energy ray image datasets and high-energy ray image datasets are acquired respectively, and a correction is performed at the edge computing node; Based on the low-energy ray image dataset and high-energy ray image dataset after one correction, an attenuation inversion function is established in the edge computing node to perform secondary correction on the boundary region between bone and soft tissue. Based on the low-energy ray image dataset and high-energy ray image dataset after secondary correction, the energy ratio sequence is calculated, and the energy ratio is dynamically smoothed in the edge computing node to obtain a stable ratio image. Based on the stable ratio image, the bone tissue image and soft tissue image are separated, and the bone mineral density value is calculated based on the separated bone tissue image; The step of performing a correction at the edge computing node includes: acquiring multiple frames of low-energy ray images and high-energy ray images, adding a time index to each frame; for each pixel, calculating the difference sequence between the gray value and the mean gray value of the same pixel's neighborhood in consecutive frames; if the absolute difference in the difference sequence exceeds the standard deviation of the neighborhood difference for more than one-third of the frames, it is marked as a defective pixel; when the proportion of defective pixels appearing in the total number of frames exceeds a preset proportion, it is determined to be a high-level defective pixel, otherwise it is a low-level defective pixel; for high-level defective pixels, neighborhood block fitting is used for repair, and the original pixel is replaced; for low-level defective pixels, gray values ​​are generated by interpolation along the row and column directions, and the original pixel is replaced; the repaired pixel values ​​are directly used to correct the low-energy ray image dataset and the high-energy ray image dataset. The step of establishing the attenuation inversion function in the edge computing node includes: calculating the horizontal and vertical neighborhood gradients for each pixel in the first-corrected low-energy and high-energy images, identifying candidate abnormal pixels at the boundary; and calculating the energy ratio of each pixel in the first-corrected low-energy ray image and the high-energy ray image. ; Calculate the average energy ratio of each pixel within a defined neighborhood. Based on the local gradient direction information, the energy ratio difference and the local gradient direction are combined to form a pixel-level attenuation inversion function: ； ； in, These are pixel-level attenuation inversion function values; This represents the energy ratio difference. The pixel gradient direction represents the angle between the direction of grayscale change and the horizontal direction. These are the direction-weighted coefficients; The secondary correction of the boundary region between bone and soft tissue includes: adjusting the boundary candidate anomalous pixels based on the pixel-level attenuation inversion function value so that the difference between the energy ratio of the boundary candidate anomalous pixels and the neighborhood reference ratio is less than a preset difference threshold; performing a weighted average correction along the neighborhood gradient direction and the gray level of the neighboring pixels in the principal gradient direction of each boundary candidate anomalous pixel to maintain the gray level continuity of the bone and soft tissue boundary; if there are boundary candidate anomalous pixels in the continuous boundary candidate anomalous pixels whose energy ratio changes beyond the neighborhood standard deviation after weighted average correction, then the gray level is re-corrected to the mean gray level of the neighboring pixels to obtain the low-energy ray image dataset and high-energy ray image dataset after secondary correction.

2. The dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in claim 1, characterized in that: The determination of the boundary candidate abnormal pixels includes: In the edge computing node, the changes in the horizontal and vertical neighboring pixel values ​​are compared pixel by pixel for low-energy ray images and high-energy ray images. When a pixel changes in the opposite direction to the pixel value of its horizontal or vertical neighbors and the magnitude of the change exceeds a preset change ratio of the average change of the neighbors, it is marked as a boundary candidate abnormal pixel.

3. The dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in claim 2, characterized in that: The grayscale is corrected to the mean grayscale value of neighboring pixels using the following calculation formula: ； in, The pixel gray value is the result of correction to the mean gray value of neighboring pixels; The average grayscale value of the neighboring pixels; The pixel grayscale value is corrected by weighted average. For re-correction factors; It is a set of local neighborhood pixels.

4. The dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in claim 3, characterized in that: The dynamic smoothing of the energy ratio in the edge computing node includes: If a certain energy ratio in the energy ratio sequence does not change in the same direction as the neighboring pixels in the same row or column, it is marked as a candidate abnormal pixel; By combining the energy ratio changes of the neighborhood in the same direction, the energy ratio of candidate abnormal pixels is adjusted to the extended value of the neighborhood trend; Check the trend consistency of each candidate anomalous pixel in the horizontal and vertical directions, fine-tune candidate anomalous pixels whose trend direction deviates from the neighborhood pattern, ensure the grayscale structure of the bone boundary and soft tissue transition area is stable, and output a stable ratio image for each frame.

5. The dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in claim 4, characterized in that: The separated bone tissue images and soft tissue images include: For each pixel in the stable ratio image, calculate the difference between the energy ratio and the median difference between the energy ratio and the ratios of neighboring pixels in the horizontal and vertical directions. When the energy ratio is higher than the median difference of the neighboring pixel ratios and exceeds a preset difference threshold, it is determined to be a bone tissue pixel; the remaining pixels are determined to be soft tissue pixels. For pixels in the boundary region between bone tissue pixels and soft tissue pixels, determine whether the energy ratio is consistent with the neighborhood direction of the bone tissue pixels. If they are consistent, they are classified as bone tissue; otherwise, they are classified as soft tissue, ensuring the continuity of the bone tissue boundary. Based on the pixel determination results, bone tissue pixels are used to form bone tissue images, and soft tissue pixels are used to form soft tissue images.

6. The dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in claim 5, characterized in that: The calculation of bone mineral density values ​​based on the separated bone tissue images includes: For each pixel in the bone tissue image, the energy ratio in the stable ratio image is extracted, and the difference in energy ratio with the neighboring pixels in the horizontal and vertical directions is statistically analyzed to form a local bone resorption index. By combining the local bone resorption index with the pre-calibrated low-energy to high-energy ratio and the relationship with bone mineral density, the local bone resorption index of each pixel is mapped to a specific bone mineral density value, forming a pixel-level bone mineral density value. The bone mineral density values ​​of all bone tissue pixels are combined to generate a bone mineral density map, while the grayscale of soft tissue pixels is set to zero.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the dual-energy X-ray bone mineral density assessment method based on a flat panel detector as described in any one of claims 1 to 6.

Images