A dual-energy bone density calculation method, system and storage medium

Abstract

The application provides a dual-energy bone density calculation method, system and storage medium, the method comprising: taking an empty exposure image by empty exposure shooting through the dual-energy X-ray exposure parameter optimized by a multi-layer neural network OptimalNet; converting the empty exposure image and the image combined with different phantoms into an attenuation image to perform beam hardening correction to obtain a linear attenuation image, and performing dual-energy subtraction and scale total variation processing to obtain a soft tissue thickness image; performing a bone tissue image subtraction operation on the linear attenuation image according to the soft tissue thickness image to obtain a bone tissue region through region growing, performing bone tissue signal statistical analysis on the bone tissue region to obtain bone mineral content and bone mineral area, and performing ratio operation on the bone mineral content and the bone mineral area to obtain bone density. The application solves the problem that the measurement accuracy of bone density is not disturbed by the fatness of the photographer and realizes bone density measurement of different parts.

Description

Technical Field

[0001] This invention relates to the field of bone mineral density measurement technology, specifically to a dual-energy bone mineral density calculation method, system, and storage medium. Background Technology

[0002] BMD: Osteoporosis is characterized by a decrease in bone mass, and the main indicator reflecting bone mass is bone mineral density (BMD), which is an important indicator of bone strength. It is expressed in grams per square centimeter and is an absolute value.

[0003] As people age, most experience bone loss. If this bone loss is severe, osteoporosis can develop. Osteoporosis not only increases the risk of fractures but also causes various complications, seriously threatening human health. Therefore, the prevention and diagnosis of osteoporosis are crucial. Bone mineral density measurement is currently an important method for evaluating bone loss and diagnosing osteoporosis.

[0004] Individual differences exist among users, which can affect bone mineral density measurements to varying degrees, such as body weight. Furthermore, linear array detectors are mostly dedicated to specific devices and cannot be used with a wide range of DR equipment, nor can they meet the requirements for bone mineral density measurements in different locations, such as the forelimbs, lumbar spine, and femur. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a dual-energy bone mineral density calculation method, system and storage medium to address the shortcomings of the prior art.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A dual-energy bone mineral density calculation method, comprising the following steps:

[0007] Grayscale images are obtained by capturing images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters. The grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image are then input into a multilayer neural network OptimalNet to select dual-energy X-ray exposure parameters.

[0008] Aerial exposure imaging is performed according to the dual-energy X-ray exposure parameters to obtain an aerial exposure image. The pre-constructed dual-head deep convolutional neural network MattNet is trained using the grayscale image and the aerial exposure image. The captured target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding aerial exposure image in real time.

[0009] The empty exposure image and images of different phantom combinations are converted into attenuated images by a pre-constructed deep convolutional neural network CorrNet. The attenuated images are then subjected to beam hardening correction to obtain linear attenuated images.

[0010] The target image of soft tissue thickness is obtained by performing a subtraction operation on the linear attenuation image. The target image of soft tissue thickness is then subjected to total variational scaling, and the solution of the soft tissue image is obtained by using a bisection iterative method.

[0011] The bone tissue image is obtained by subtracting the linear attenuation image from the soft tissue thickness image. The bone tissue image is then grown to obtain the bone tissue region. Statistical analysis of the bone tissue signal is performed on the bone tissue region to obtain the bone mineral content and the area of ​​the bone minerals. The ratio of the bone mineral content and the area of ​​the bone minerals is then calculated to obtain the bone density.

[0012] Based on the above technical solution, the present invention can be further improved as follows:

[0013] Furthermore, grayscale images are obtained by capturing images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters. The grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image are then input into a multilayer neural network OptimalNet to optimize the dual-energy X-ray exposure parameters. This process includes:

[0014] Pre-collect different voltages (kV) i Different currents mAs n PMMA of equivalent human soft tissue of different thicknesses i Equivalent human bone tissue AL of different thicknesses k and different grayscale signal-to-noise ratios (SNR) n Construct the training sample set {kv i mAs n PMMA i ,AL k ,SNRn};

[0015] Construct a multi-layer neural network OptimalNet, and use {kv i mAs n PMMA i ,AL k The OptimalNet model is trained using the training sample set {SNRn} to obtain the trained multilayer neural network OptimalNet.

[0016] Combining different phantoms {PMMA i ,AL k}, different voltages {kV i}, different currents {mAs n The input is fed into the trained multilayer neural network OptimalNet, and the output is the SNR value;

[0017] Based on the preset SNR value, from all phantom combinations {PMMA i ,AL k} of {kv estimated mAs estimated} Select the largest kv from the combinations estimated kV as a dose parameter for high-energy imaging H and the smallest kv estimated kV as a dose parameter for high-energy imaging L ; Determine the kv for high-energy shooting H Then, select the smallest mAs under that kv. estimated As the optimal dose parameter mAs H ; Determine the kV for low-energy shooting L Then, select the smallest mAs under that kv. estimated As the optimal dose parameter mAs L .

[0018] Further, the process of performing a void exposure image according to the dual-energy X-ray exposure parameters to obtain a void exposure image, and training a pre-constructed dual-head deep convolutional neural network MattNet using the grayscale image and the void exposure image, and inputting the captured target grayscale image into the trained dual-head deep convolutional neural network MattNet to generate the corresponding void exposure image in real time, includes:

[0019] Blank, the initial blank exposure image pre-acquired i Image of the phantom i And binary image MattImage i As training samples;

[0020] Construct a deep convolutional neural network MattNet, and then... i MattImage i Blank i The MattNet model is trained using these samples to obtain the trained deep convolutional neural network MattNet.

[0021] The target grayscale image (Image) of the object is input into the trained deep convolutional neural network MattNet to generate the corresponding blank exposure image (Blank) in real time. i .

[0022] Furthermore, the process of converting the unexposed image and images of different phantom combinations into attenuated images using a pre-constructed deep convolutional neural network CorrNet, and then performing beam hardening correction on the attenuated images to obtain linear attenuated images includes:

[0023] Pre-acquisition of initial low-energy decay images Atten L High-energy decay image Atten H Low-energy linear decay image Atten L * And high-energy linear decay image Atten H * ;

[0024] Construct a deep convolutional neural network CorrNet, and then {Atten L Atten H Atten L * Atten H * The CorrNet ​​model is trained using these samples to obtain the trained CorrNet ​​model.

[0025] Atten low-energy decay image L and high-energy decay image Atten H Input the trained CorrNet ​​model and output a low-energy linearly decaying image Atten L * And high-energy linear decay image Atten H * .

[0026] Further, the process of performing a subtraction operation on the linearly attenuated image to obtain a target image of soft tissue thickness, performing total variational scaling on the target image of soft tissue thickness, and using a bisection iterative method to obtain the solution of the soft tissue image to obtain the soft tissue thickness image includes:

[0027] According to the low-energy linear decay image Atten L * And high-energy linear decay image Atten H * To perform a subtraction operation, the specific steps are as follows:

[0028] right and Solving the simultaneous equations yields human soft tissue of different thicknesses.

[0029]

[0030] in, The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue;

[0031] according to and The target image for calculating the soft tissue thickness using initial values:

[0032]

[0033] Wherein, SoftImage is the target image of soft tissue thickness. The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human bone tissue;

[0034] The process of performing scale total variation processing on the target image of soft tissue thickness is as follows:

[0035] ScaleTV=∑|(gauss(SoftImage,r0)-gauss(SoftImage,r1)|,

[0036] Where ScaleTV represents the total variation coefficients, SoftImage represents the target image of soft tissue thickness, gauss represents the Gaussian filter function with scale r, r0 represents the small scale, and r1 represents the large scale.

[0037] Will The ratio ф is used for binary iterative search to select the optimal ф. optimal Minimize the total variation ScaleTV of the target image SoftImage; to obtain the soft tissue image.

[0038]

[0039] Where SoftImage is the target image of soft tissue thickness, and ф is the ratio of the attenuation coefficients of low-energy equivalent human bone tissue. optimal The ratio of the attenuation coefficients of low- to high-energy equivalent human bone tissue that minimizes the total variation coefficients of the scale, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image.

[0040] Further, the process of obtaining a bone tissue image by performing a subtraction operation on the linear attenuation image based on the soft tissue thickness image, obtaining a bone tissue region by performing region growing on the bone tissue image, obtaining bone mineral content and bone mineral area by performing statistical analysis on the bone tissue region, and obtaining bone density by calculating the ratio of bone mineral content and bone mineral area includes:

[0041] The linear attenuation image is subtracted based on the soft tissue thickness image, specifically as follows:

[0042] right and By solving the simultaneous equations, we can obtain human bone tissue of different thicknesses.

[0043]

[0044] Among them, Atten L * For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy bone tissue lines. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue;

[0045] AL k Multiplying the corresponding bone mineral density coefficient yields a bone mineral content map. BMC ;

[0046] The binary image obtained after binarizing the bone mineral content map is subjected to erosion, dilation, and region growing to obtain closed bone tissue regions. A threshold is set for the area of ​​these closed bone tissue regions, and they are then filtered to obtain the final binary image, "BinaryImage". BMC ;

[0047] The effective bone mineral content (BMC) is calculated based on the bone mineral content calculation formula, which is as follows:

[0048]

[0049] BMC represents bone mineral content. BMC Image showing bone mineral content. (BinaryImage) BMC The image is a binary image, where (x, y) is a two-dimensional image. BMC coordinates

[0050] The area of ​​bone mineral deposits is calculated based on the formula for calculating bone mineral area, thus obtaining the bone mineral area (Area). BMC The formula for calculating the bone mineral area is:

[0051] Area BMC =∑BinaryImage BMC (x,y),

[0052] Among them, Area BMC Let (x, y) be the area of ​​bone mineral deposits, and (x, y) be the two-dimensional image. BMC coordinates

[0053] Based on the bone mineral density formula, the bone mineral content (BMC) and the area of ​​bone mineral density are combined. BMC Bone mineral density is obtained by performing a ratio calculation. The formula for bone mineral density is:

[0054] BMD = BMC / Area BMC ,

[0055] Where BMD is bone mineral density, BMC is bone mineral content, and Area BMC This represents the area of ​​bone mineral deposits.

[0056] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a dual-energy bone mineral density calculation system, including a dual-energy X-ray exposure parameter optimization module, a dual-energy X-ray real-time spatial exposure image generation module, a dual-energy X-ray beam hardening correction module, a soft tissue thickness image generation module, and a bone mineral density statistical analysis and calculation module;

[0057] The dual-energy X-ray exposure parameter optimization module is used to capture grayscale images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters, and input the grayscale images, set acquisition parameters, and signal-to-noise ratio (SNR) of the target image into a multilayer neural network OptimalNet to optimize the dual-energy X-ray exposure parameters.

[0058] The dual-energy X-ray real-time air exposure image generation module is used to perform air exposure shooting according to the dual-energy X-ray exposure parameters to obtain an air exposure image, and to train a pre-constructed dual-head deep convolutional neural network MattNet using the grayscale image and the air exposure image. The target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding air exposure image in real time.

[0059] The dual-energy X-ray beam hardening correction module is used to convert the empty exposure image and the images of different phantom combinations into attenuated images through a pre-constructed deep convolutional neural network CorrNet, and to perform beam hardening correction on the attenuated images to obtain linear attenuated images.

[0060] The soft tissue thickness image generation module is used to perform a subtraction operation on the linear attenuation image to obtain a target image of soft tissue thickness, perform scale total variation processing on the target image of soft tissue thickness, and use the bisection iterative method to obtain the solution of the soft tissue image to obtain the soft tissue thickness image.

[0061] The bone density statistical analysis and calculation module is used to perform a subtraction operation on the linear attenuation image based on the soft tissue thickness image to obtain a bone tissue image, perform region growing on the bone tissue image to obtain a bone tissue region, perform bone tissue signal statistical analysis on the bone tissue region to obtain bone mineral content and bone mineral area, and calculate the ratio between bone mineral content and bone mineral area to obtain bone density.

[0062] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a dual-energy bone mineral density calculation system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the dual-energy bone mineral density calculation method as described above.

[0063] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the dual-energy bone density calculation method as described above.

[0064] The beneficial effects of this invention are as follows: A blank exposure image is obtained by using dual-energy X-ray exposure parameters optimized through a multi-layer neural network OptimalNet; the blank exposure image and images of different phantom combinations are converted into attenuated images, which are then subjected to beam hardening correction to obtain linearly attenuated images. These linearly attenuated images are then subjected to dual-energy subtraction and total variational scaling to obtain a soft tissue thickness image; based on the soft tissue thickness image, the linearly attenuated image is subtracted to obtain a bone tissue image, which is then subjected to region growing to obtain a bone tissue region. Statistical analysis of the bone tissue signals in the bone tissue region yields the bone mineral content and area. The ratio of the bone mineral content to the area of ​​the bone mineral is then calculated to obtain bone mineral density. This invention achieves the effect that the measurement accuracy of bone mineral density is not affected by the size of the person being photographed, and it enables bone mineral density measurement of different body parts. It has good repeatability, stable operation, is easy to use, and has stronger compatibility. Attached Figure Description

[0065] Figure 1 A schematic flowchart of the dual-energy bone mineral density calculation method provided in an embodiment of the present invention;

[0066] Figure 2 This is a block diagram of a dual-energy bone mineral density calculation system provided in an embodiment of the present invention. Detailed Implementation

[0067] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0068] like Figure 1 As shown in the figure, an embodiment of the present invention provides a dual-energy bone mineral density calculation method, which includes the following steps:

[0069] Grayscale images are obtained by capturing images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters. The grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image are then input into a multilayer neural network OptimalNet to select dual-energy X-ray exposure parameters.

[0070] Aerial exposure imaging is performed according to the dual-energy X-ray exposure parameters to obtain an aerial exposure image. The pre-constructed dual-head deep convolutional neural network MattNet is trained using the grayscale image and the aerial exposure image. The captured target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding aerial exposure image in real time.

[0071] The empty exposure image and images of different phantom combinations are converted into attenuated images by a pre-constructed deep convolutional neural network CorrNet. The attenuated images are then subjected to beam hardening correction to obtain linear attenuated images.

[0072] The target image of soft tissue thickness is obtained by performing a subtraction operation on the linear attenuation image. The target image of soft tissue thickness is then subjected to total variational scaling, and the solution of the soft tissue image is obtained by using a bisection iterative method.

[0073] The bone tissue image is obtained by subtracting the linear attenuation image from the soft tissue thickness image. The bone tissue image is then grown to obtain the bone tissue region. Statistical analysis of the bone tissue signal is performed on the bone tissue region to obtain the bone mineral content and the area of ​​the bone minerals. The ratio of the bone mineral content and the area of ​​the bone minerals is then calculated to obtain the bone density.

[0074] In the above embodiments, an empty exposure image is obtained by using dual-energy X-ray exposure parameters optimized by the multi-layer neural network OptimalNet. The empty exposure image and images of different phantom combinations are converted into attenuated images, which are then subjected to beam hardening correction. The linearly attenuated images obtained are then subjected to dual-energy subtraction and total variational scaling to obtain a soft tissue thickness image. Based on the soft tissue thickness image, a bone tissue image is obtained by performing subtraction on the linearly attenuated image. Region growing is then performed on the bone tissue image to obtain a bone tissue region. Statistical analysis of the bone tissue signals in the bone tissue region yields the bone mineral content and area. The ratio of the bone mineral content to the area of ​​the bone mineral is then used to calculate bone mineral density. This invention achieves bone mineral density measurement accuracy unaffected by the subject's weight, enables bone mineral density measurement of different body parts, exhibits good repeatability, stable operation, ease of use, and enhanced compatibility.

[0075] Optionally, the process of capturing grayscale images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters, and then inputting the grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image into a multilayer neural network OptimalNet to optimize dual-energy X-ray exposure parameters includes:

[0076] By setting different voltage values ​​kv i Current mAs of different values n PMMA of equivalent human soft tissue of different thicknesses i Equivalent human bone tissue AL of different thicknesses k And different values ​​of grayscale signal-to-noise ratio (SNR) n Construct the training sample set {kv i mAs n PMMA i ,AL k ,SNRn};

[0077] Construct a multi-layer neural network OptimalNet using a training sample set {kv} i mAsn PMMA i AL k , SNRn} train the OptimalNet model to obtain the trained multilayer neural network OptimalNet;

[0078] Combining different phantoms {PMMA i ,AL k The voltage {kV} i The current {mAs} n The input is fed into the trained multilayer neural network OptimalNet to predict the SNR value;

[0079] It should be understood that the acquired dual-energy X-ray exposure parameters are kV and mAs values. The specific acquisition parameters are set as follows: based on the kV values ​​supported by the high-voltage generator, a range is set every 5 kV to establish a set of available kV values ​​{kV}. i Multiple 1cm thick PMMA phantoms are stacked and combined to simulate equivalent human soft tissue of different thicknesses. i PMMA is made of polymethyl methacrylate; multiple 1mm thick AL phantoms are stacked and combined to simulate equivalent human bone tissue of different thicknesses. k}, AL represents aluminum; a fixed kv i {mAs supported by high voltage generator n}, for combinations of different phantoms {PMMA i ,AL k} Take pictures and obtain image sets {Image n}, for the image Image n Threshold segmentation is performed, and the mean grayscale value of the phantom region is calculated. n Standard deviation Std n The gray-scale signal-to-noise ratio (SNR) of the phantom region was calculated. n =Mean n / Std n .

[0080] In the above embodiments, the exposure parameters for X-ray imaging can be selected quickly and accurately, improving the efficiency and accuracy of bone density measurement.

[0081] Optionally, an empty exposure image is obtained by performing an empty exposure image according to the dual-energy X-ray exposure parameters. A pre-constructed dual-head deep convolutional neural network, MattNet, is then trained using the grayscale image and the empty exposure image. The process of inputting the captured target grayscale image into the trained MattNet to generate the corresponding empty exposure image in real time includes:

[0082] Blank, the initial blank exposure image pre-acquired i Image of the phantom i And binary image MattImage i , obtain the collection {Image i MattImage i Blank i};

[0083] Construct a deep convolutional neural network MattNet, which combines the set {Image i MattImage i Blank i The MattNet model is trained using these samples to obtain the trained deep convolutional neural network MattNet.

[0084] The target grayscale image (Image) of the object is input into the trained deep convolutional neural network MattNet to generate the corresponding blank exposure image (Blank) in real time. i It also generates a binary image, MattImage.

[0085] The above embodiments effectively avoid the problem of X-ray dose differences caused by hardware such as high-voltage generators, thus saving shooting costs.

[0086] Optionally, the process of converting the unexposed image and images of different phantom combinations into attenuated images using a pre-constructed deep convolutional neural network CorrNet, and then performing beam hardening correction on the attenuated images to obtain linear attenuated images includes:

[0087] Pre-acquisition of initial low-energy decay images Atten L High-energy decay image Atten H Low-energy linear decay image Atten L * And high-energy linear decay image Atten H * , thus obtaining the set {Atten L Atten H Atten L * Atten H *};

[0088] Construct a deep convolutional neural network CorrNet, and combine the set {Atten L Atten H Atten L * Atten H* The CorrNet ​​model is trained using these samples to obtain the trained CorrNet ​​model.

[0089] Atten low-energy decay image L and high-energy decay image Atten H Inputting the trained CorrNet ​​model yields linear decay images, including low-energy linear decay images (Atten). L * And high-energy linear decay image Atten H * .

[0090] It should be understood that the process of converting an empty exposure image and images of different phantom combinations into a decayed image is as follows:

[0091] Based on the preferred low-energy imaging dose parameter kv L and mAs L High-energy imaging dose parameters (kV) H and mAs H First, perform an aerial exposure shot to obtain a low-energy aerial exposure image I. 0L High-energy space exposure image I 0H ; Combinations of different phantoms {PMMA i ,AL k}Shooting was performed to obtain low-energy images I. L With high-energy image I H Converting grayscale images into low-energy decay images and high-energy decay images: Atten L =ln(I 0L / I L Atten H =ln(I 0H / I H ); Determination of PMMA and AL at kV L ,kv H Linear attenuation coefficient (unit: cm) -1 ), respectively According to the combination {PMMA i ,AL k Construct a low-energy target decay image: And high-energy target decay images:

[0092] In the above embodiments, the effects of dual-energy beam sclerosis are reduced, and the accuracy of bone mineral density measurement is improved.

[0093] Optionally, the process of performing a subtraction operation on the linearly attenuated image to obtain a target image of soft tissue thickness, performing total variational scaling on the target image of soft tissue thickness, and using a bisection iterative method to obtain the solution of the soft tissue image to obtain the soft tissue thickness image includes:

[0094] According to the low-energy linear decay image Atten L * And high-energy linear decay image Atten H * To perform a subtraction operation, the specific steps are as follows:

[0095] right and Solving the simultaneous equations yields human soft tissue of different thicknesses.

[0096]

[0097] in, The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue;

[0098] according to and The target image for calculating the soft tissue thickness using initial values:

[0099]

[0100] Wherein, SoftImage is the target image of soft tissue thickness. The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human bone tissue;

[0101] The target image of soft tissue thickness is subjected to total variational scaling, specifically:

[0102] ScaleTV=∑|(gauss(SoftImage,r0)-gauss(SoftImage,r1)|,

[0103] Where ScaleTV represents the total variation coefficients of the scale, gauss represents the Gaussian filter function with scale r, r0 represents the small scale, and r1 represents the large scale;

[0104] Will The ratio ф is used for binary iterative search to select the optimal ф. optimal Minimize the total variation ScaleTV of the target image SoftImage; to obtain the soft tissue image.

[0105]

[0106] Where SoftImage is the target image of soft tissue thickness, and ф is the ratio of the attenuation coefficients of low-energy equivalent human bone tissue. optimal The ratio of the attenuation coefficients of low- to high-energy equivalent human bone tissue that minimizes the total variation coefficients of the scale, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image.

[0107] Beneficial effect: By using an iterative method to obtain images of soft tissue thickness, the accuracy of bone mineral density measurement is improved.

[0108] Optionally, the process of obtaining a bone tissue image by performing a subtraction operation on a linear attenuation image based on the soft tissue thickness image, obtaining a bone tissue region by performing region growing on the bone tissue image, obtaining bone mineral content and bone mineral area by performing statistical analysis on the bone tissue region, and obtaining bone density by calculating the ratio of bone mineral content and bone mineral area includes:

[0109] The linear attenuation image is subtracted based on the soft tissue thickness image, specifically as follows:

[0110] right and Solving the simultaneous equations yields equivalent human bone tissue of different thicknesses.

[0111]

[0112] Among them, Atten L *For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue;

[0113] AL k Multiplying the corresponding bone mineral density coefficient yields a bone mineral content map. BMC ;

[0114] The binary image obtained after binarizing the bone mineral content map is subjected to erosion, dilation, and region growing to obtain closed bone tissue regions. A threshold is set for the area of ​​these closed bone tissue regions, and they are then filtered to obtain the final binary image, "BinaryImage". BMC ;

[0115] The effective bone mineral content (BMC) is calculated based on the bone mineral content calculation formula, which is as follows:

[0116]

[0117] BMC represents bone mineral content. BMC Image showing bone mineral content. (BinaryImage) BMC The image is a binary image, where (x, y) is a two-dimensional image. BMC coordinates

[0118] The area of ​​bone mineral deposits is calculated based on the formula for calculating bone mineral area, thus obtaining the bone mineral area (Area). BMC The formula for calculating the bone mineral area is:

[0119] Area BMC =∑BinaryImage BMC (x,y),

[0120] Among them, Area BMC Let (x, y) be the area of ​​bone mineral deposits, and (x, y) be the two-dimensional image. BMC coordinates

[0121] Based on the bone mineral density formula, the bone mineral content (BMC) and the area of ​​bone mineral density are combined. BMCBone mineral density is obtained by performing a ratio calculation. The formula for bone mineral density is:

[0122] BMD = BMC / Area BMC ,

[0123] Where BMD is bone mineral density, BMC is bone mineral content, and Area BMC This represents the area of ​​bone mineral deposits.

[0124] Beneficial effects: Among the possible solutions to the equation, the solution with the optimal matching constraint is found, thus enabling accurate calculation of the equivalent bone mineral content of bone tissue; moreover, the module adopts an iterative convergence method, and when encountering a solution mismatch problem caused by X-ray dose error, it can also provide the optimal solution under the current conditions, improving the robustness of the measurement system.

[0125] It should be understood that before binarizing the bone mineral content map, pixels with a value greater than 0 are considered to be areas containing bone minerals.

[0126] It should be understood that the erosion process is to change the grayscale value of all pixels within a specific radius to 0, and the dilation process is to change the grayscale value of all pixels within a specific radius to the maximum value, thereby eliminating noise interference; region growing refers to starting from a certain pixel and gradually adding neighboring pixels according to certain criteria, and terminating region growing when certain conditions are met.

[0127] It should be understood that the stated Linear attenuation coefficient is The value closest to ф optimal Obtained from below.

[0128] like Figure 2 As shown, the present invention provides a dual-energy bone mineral density calculation system, including a dual-energy X-ray exposure parameter optimization module, a dual-energy X-ray real-time spatial exposure image generation module, a dual-energy X-ray beam hardening correction module, a soft tissue thickness image generation module, and a bone mineral density statistical analysis and calculation module.

[0129] The dual-energy X-ray exposure parameter optimization module is used to capture grayscale images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters, and input the grayscale images, set acquisition parameters, and signal-to-noise ratio (SNR) of the target image into a multilayer neural network OptimalNet to optimize the dual-energy X-ray exposure parameters.

[0130] The dual-energy X-ray real-time air exposure image generation module is used to perform air exposure shooting according to the dual-energy X-ray exposure parameters to obtain an air exposure image, and to train a pre-constructed dual-head deep convolutional neural network MattNet using the grayscale image and the air exposure image. The target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding air exposure image in real time.

[0131] The dual-energy X-ray beam hardening correction module is used to convert the empty exposure image and the images of different phantom combinations into attenuated images through a pre-constructed deep convolutional neural network CorrNet, and to perform beam hardening correction on the attenuated images to obtain linear attenuated images.

[0132] The soft tissue thickness image generation module is used to perform a subtraction operation on the linear attenuation image to obtain a target image of soft tissue thickness, perform scale total variation processing on the target image of soft tissue thickness, and use the bisection iterative method to obtain the solution of the soft tissue image to obtain the soft tissue thickness image.

[0133] The bone density statistical analysis and calculation module is used to perform a subtraction operation on the linear attenuation image based on the soft tissue thickness image to obtain a bone tissue image, perform region growing on the bone tissue image to obtain a bone tissue region, perform bone tissue signal statistical analysis on the bone tissue region to obtain bone mineral content and bone mineral area, and calculate the ratio between bone mineral content and bone mineral area to obtain bone density.

[0134] This invention provides a dual-energy bone mineral density calculation system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the dual-energy bone mineral density calculation method as described above.

[0135] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the dual-energy bone density calculation method described above.

[0136] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0139] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0140] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0141] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This is understood to mean that the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A dual-energy bone mineral density calculation method, characterized in that, Includes the following steps: Grayscale images are obtained by capturing images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters. The grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image are then input into a multilayer neural network OptimalNet to select dual-energy X-ray exposure parameters. Aerial exposure imaging is performed according to the dual-energy X-ray exposure parameters to obtain an aerial exposure image. The pre-constructed dual-head deep convolutional neural network MattNet is trained using the grayscale image and the aerial exposure image. The captured target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding aerial exposure image in real time. The empty exposure image and images of different phantom combinations are converted into attenuated images by a pre-constructed deep convolutional neural network CorrNet. The attenuated images are then subjected to beam hardening correction to obtain linear attenuated images. The target image of soft tissue thickness is obtained by performing a subtraction operation on the linear attenuation image. The target image of soft tissue thickness is then subjected to total variational scaling, and the solution of the soft tissue image is obtained by using a bisection iterative method. The bone tissue image is obtained by subtracting the linear attenuation image from the soft tissue thickness image. The bone tissue image is then grown to obtain the bone tissue region. Statistical analysis of the bone tissue signal is performed on the bone tissue region to obtain the bone mineral content and the area of ​​the bone minerals. The ratio of the bone mineral content and the area of ​​the bone minerals is then calculated to obtain the bone density.

2. The dual-energy bone mineral density calculation method according to claim 1, characterized in that, The process of capturing grayscale images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters, and then inputting the grayscale images, the set acquisition parameters, and the signal-to-noise ratio (SNR) of the target image into a multilayer neural network OptimalNet to optimize dual-energy X-ray exposure parameters includes: By setting different voltage values ​​kv i Current mAs of different values n PMMA of equivalent human soft tissue of different thicknesses i Equivalent human bone tissue AL of different thicknesses k And different values ​​of grayscale signal-to-noise ratio (SNR) n Construct the training sample set {kv i mAs n PMMA i ,AL k ,SNRn}; Construct a multi-layer neural network OptimalNet using a training sample set {kv} i mAs n PMMA i AL k , SNRn} train the OptimalNet model to obtain the trained multilayer neural network OptimalNet; Combining different phantoms {PMMA i AL k The voltage {kV} i The current {mAs} n The input is fed into the trained multilayer neural network OptimalNet to predict the SNR value; Based on the preset SNR value, from all phantom combinations {PMMA i ,AL k } of {kv estimated mAs estimated } Select the largest kv from the combinations estimated kV as a dose parameter for high-energy imaging H and the smallest kv estimated kV as a dose parameter for high-energy imaging L ; Determine the kv for high-energy shooting H Then, select the smallest mAs under that kv. estimated As the optimal dose parameter mAs H ; Determine the kV for low-energy shooting L Then, select the smallest mAs under that kv. estimated As the optimal dose parameter mAs L .

3. The dual-energy bone mineral density calculation method according to claim 1, characterized in that, Aerial exposure imaging is performed according to the aforementioned dual-energy X-ray exposure parameters to obtain an aerial exposure image. The pre-constructed dual-head deep convolutional neural network MattNet is then trained using the grayscale image and the aerial exposure image. The process of inputting the captured target grayscale image into the trained MattNet to generate the corresponding aerial exposure image in real time includes: Blank, the initial blank exposure image pre-acquired i Image of the phantom i And binary image MattImage i , obtain the set {Image i MattImage i Blank i }; Construct a deep convolutional neural network MattNet, which combines the set {Image i MattImage i Blank i The MattNet model is trained using these samples to obtain the trained deep convolutional neural network MattNet. The target grayscale image (Image) of the object is input into the trained deep convolutional neural network MattNet to generate the corresponding blank exposure image (Blank) in real time. i .

4. The dual-energy bone mineral density calculation method according to claim 1, characterized in that, The process of converting the unexposed image and images of different phantom combinations into attenuated images using a pre-constructed deep convolutional neural network CorrNet, and then performing beam hardening correction on the attenuated images to obtain linearly attenuated images includes: Pre-acquisition of initial low-energy decay images Atten L High-energy decay image Atten H Low-energy linear decay image Atten L * And high-energy linear decay image Atten H * , thus obtaining the set {Atten L Atten H Atten L * Atten H * }; Construct a deep convolutional neural network CorrNet, and combine the set {Atten L Atten H Atten L * Atten H * The CorrNet ​​model is trained using these samples to obtain the trained CorrNet ​​model. Atten low-energy decay image L and high-energy decay image Atten H Inputting the trained CorrNet ​​model yields linear decay images, including low-energy linear decay images (Atten). L * And high-energy linear decay image Atten H * .

5. The dual-energy bone mineral density calculation method according to claim 4, characterized in that, The process of obtaining a target image of soft tissue thickness by performing a subtraction operation on the linearly attenuated image, performing total variational scaling on the target image of soft tissue thickness, and using a bisection iterative method to obtain the solution of the soft tissue image includes: According to the low-energy linear decay image Atten L * And high-energy linear decay image Atten H * To perform a subtraction operation, the specific steps are as follows: right and Solving the simultaneous equations yields equivalent human soft tissue of different thicknesses: in, The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue; according to and The target image for calculating the soft tissue thickness using initial values: Wherein, SoftImage is the target image of soft tissue thickness. The ratio of the attenuation coefficients of low-energy and high-energy equivalent human bone tissue, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human bone tissue; The process of performing scale total variation processing on the target image of soft tissue thickness is as follows: ScaleTV=Σ|(gauss(SoftImage,r0)-gauss(SoftImage,r1)|, Where ScaleTV represents the total variation coefficients, SoftImage represents the target image of soft tissue thickness, gauss represents the Gaussian filter function with scale r, r0 represents the small scale, and r1 represents the large scale; Will The ratio ф is used for binary iterative search to select the optimal ф. optimal To obtain a soft tissue image, minimize the total scale variation ScaleTV of the target image SoftImage. Where SoftImage is the target image of soft tissue thickness, and ф is the ratio of the attenuation coefficients of low-energy equivalent human bone tissue. optimal The ratio of the attenuation coefficients of low- to high-energy equivalent human bone tissue that minimizes the total variation coefficients of the scale, Atten L * For low-energy linear decay images, Atten H * This is a high-energy linear decay image.

6. The dual-energy bone mineral density calculation method according to claim 5, characterized in that, The process of obtaining a bone tissue image by subtracting a linear attenuation image from a soft tissue thickness image, performing region growing on the bone tissue image to obtain a bone tissue region, performing statistical analysis of the bone tissue signal in the bone tissue region to obtain bone mineral content and bone mineral area, and calculating the ratio between bone mineral content and bone mineral area to obtain bone density includes: The linear attenuation image is subtracted based on the soft tissue thickness image, specifically as follows: right and Solving the simultaneous equations yields equivalent human bone tissue of different thicknesses. Among them, Atten L * For low-energy linear decay images, Atten H * For high-energy linear decay images, PMMA i For equivalent human soft tissue of different thicknesses, AL k These represent equivalent human bone tissue of varying thicknesses. The attenuation coefficient of low-energy equivalent human soft tissue lines. The attenuation coefficient of low-energy equivalent human bone tissue line. The attenuation coefficient of high-energy equivalent human soft tissue lines. The attenuation coefficient of high-energy equivalent human bone tissue; Equivalent human bone tissue AL k Multiplying the corresponding bone mineral density coefficient yields a bone mineral content map. BMC ; Image of bone mineral content BMC The binary image obtained after binarization is subjected to erosion, dilation, and region growing to obtain the bone tissue region. An area threshold is set for the bone tissue region to filter out closed bone tissue regions, resulting in the binary image (BinaryImage). BMC ; The effective bone mineral content (BMC) is calculated based on the bone mineral content calculation formula, which is as follows: BMC represents bone mineral content. BMC Image showing bone mineral content. (BinaryImage) BMC The image is a binary image, where (x, y) is a two-dimensional image. BMC coordinates The area of ​​bone mineral deposits is calculated based on the formula for calculating bone mineral area, thus obtaining the bone mineral area (Area). BMC The formula for calculating the bone mineral area is: Area BMC =∑BinaryImage BMC (x,y), Among them, Area BMC Let (x, y) be the area of ​​bone mineral deposits, and (x, y) be the two-dimensional image. BMC coordinates Based on the bone mineral density formula, the bone mineral content (BMC) and the area of ​​bone mineral density are combined. BMC Bone mineral density is obtained by performing a ratio calculation. The formula for bone mineral density is: BMD=BMC / Area BMC , Where BMD is bone mineral density, BMC is bone mineral content, and Area BMC This represents the area of ​​bone mineral deposits.

7. A dual-energy bone mineral density calculation system, characterized in that, It includes a dual-energy X-ray exposure parameter optimization module, a dual-energy X-ray real-time spatial exposure image generation module, a dual-energy X-ray beam hardening correction module, a soft tissue thickness image generation module, and a bone density statistical analysis and calculation module; The dual-energy X-ray exposure parameter optimization module is used to capture grayscale images of equivalent PMMA phantoms and equivalent AL phantoms of different thicknesses using set acquisition parameters, and input the grayscale images, set acquisition parameters, and signal-to-noise ratio (SNR) of the target image into a multilayer neural network OptimalNet to optimize the dual-energy X-ray exposure parameters. The dual-energy X-ray real-time air exposure image generation module is used to perform air exposure shooting according to the dual-energy X-ray exposure parameters to obtain an air exposure image, and to train a pre-constructed dual-head deep convolutional neural network MattNet using the grayscale image and the air exposure image. The target grayscale image is input into the trained dual-head deep convolutional neural network MattNet to generate the corresponding air exposure image in real time. The dual-energy X-ray beam hardening correction module is used to convert the empty exposure image and the images of different phantom combinations into attenuated images through a pre-constructed deep convolutional neural network CorrNet, and to perform beam hardening correction on the attenuated images to obtain linear attenuated images. The soft tissue thickness image generation module is used to perform a subtraction operation on the linear attenuation image to obtain a target image of soft tissue thickness, perform scale total variation processing on the target image of soft tissue thickness, and use the bisection iterative method to obtain the solution of the soft tissue image to obtain the soft tissue thickness image. The bone density statistical analysis and calculation module is used to perform a subtraction operation on the linear attenuation image based on the soft tissue thickness image to obtain a bone tissue image, perform region growing on the bone tissue image to obtain a bone tissue region, perform bone tissue signal statistical analysis on the bone tissue region to obtain bone mineral content and bone mineral area, and calculate the ratio between bone mineral content and bone mineral area to obtain bone density.

8. The dual-energy bone mineral density calculation system according to claim 7, characterized in that, Aerial exposure imaging is performed according to the aforementioned dual-energy X-ray exposure parameters to obtain an aerial exposure image. The pre-constructed dual-head deep convolutional neural network MattNet is then trained using the grayscale image and the aerial exposure image. The process of inputting the captured target grayscale image into the trained MattNet to generate the corresponding aerial exposure image in real time includes: Blank, the initial blank exposure image pre-acquired i Image of the phantom i And binary image MattImage i , obtain the set {Image i MattImage i Blank i }; Construct a deep convolutional neural network MattNet, which combines the set {Image i MattImage i Blank i The MattNet model is trained using these samples to obtain the trained deep convolutional neural network MattNet. The target grayscale image (Image) of the object is input into the trained deep convolutional neural network MattNet to generate the corresponding blank exposure image (Blank) in real time. i .

9. A dual-energy bone mineral density calculation system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the dual-energy bone mineral density calculation method as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the dual-energy bone mineral density calculation method as described in any one of claims 1 to 6.

Images