Bone trabecula feature extraction method based on image analysis

Abstract

The invention relates to the technical field of medical image processing, and discloses a bone trabecula feature extraction method based on image analysis, which comprises the following steps: scanning bone tissue by using X-ray imaging equipment to obtain a two-dimensional projection image of the bone tissue; the obtained two-dimensional projection image is preprocessed; segmenting the image through an edge detection segmentation method, and separating the bone trabecula from the image to obtain a bone trabecula area; constructing a deep learning model to perform morphological processing on the segmented bone trabecular region, and refining the boundary of the bone trabecular region; performing multi-time iteration feature extraction on the segmented bone trabecula region, and adjusting feature extraction parameters according to a result of previous iteration; and performing correlation analysis on the extracted bone trabecula features, and displaying the distribution and change trend of the features through a data visualization method. According to the method, the bone trabecula characteristics can be comprehensively and accurately extracted, and technical support is provided for diagnosis and research of orthopedic diseases such as osteoporosis.

Description

Technical Field

[0001] The invention relates to the technical field of medical image processing, and in particular to a method for extracting trabecular features based on image analysis. Background Art

[0002] The structural characteristics of trabecular bone are of great significance for the diagnosis of orthopedic diseases such as osteoporosis. Accurately extracting the characteristics of trabecular bone can provide an important basis for the early diagnosis and treatment of diseases. However, the existing trabecular bone feature extraction methods have certain limitations in processing complex images and extracting comprehensive features, and further optimization is needed to improve the accuracy and efficiency of feature extraction.

[0003] Existing methods for extracting trabecular bone features mainly rely on traditional image processing techniques. These methods are often affected by image noise and background interference when segmenting trabecular bone areas and extracting features, resulting in low accuracy of feature extraction. In addition, existing methods are difficult to maintain consistent segmentation results when dealing with trabecular bones of different densities and complex structures, which affects the stability of feature extraction.

[0004] In trabecular bone image processing, image segmentation is a key step in extracting features. Existing segmentation methods are prone to mis-segmentation when processing images with low contrast between trabecular bone and background, resulting in reduced accuracy of feature extraction. In addition, existing methods have low computational efficiency when processing large amounts of image data, making it difficult to meet the needs of rapid clinical diagnosis.

[0005] In recent years, deep learning technology has made significant progress in the field of image processing, especially in image segmentation and feature extraction. However, existing deep learning models still have some problems when processing trabecular bone images, such as the need for a large amount of labeled data for model training and the tendency to overfit. In addition, existing models have deficiencies in processing multi-scale features and are difficult to capture the complex structural information of trabecular bone.

[0006] Therefore, the present invention proposes a trabecular bone feature extraction method based on image analysis, which has an improved deep learning model and can better adapt to the characteristics of trabecular bone images and improve the accuracy and robustness of feature extraction. Summary of the invention

[0007] The purpose of the present invention is to solve the problems in the prior art and to propose a method for extracting trabecular features based on image analysis.

[0008] In order to achieve the above object, the present invention adopts the following technical scheme: a method for extracting trabecular features based on image analysis, comprising the following steps: Step S1, scanning the bone tissue using an X-ray imaging device to obtain a two-dimensional projection image of the bone tissue; Step S2, preprocessing the acquired two-dimensional projection image of bone tissue, including denoising, contrast enhancement, histogram equalization and normalization processing; Step S3, segmenting the image using an edge detection segmentation method to separate the trabeculae from the image to obtain a trabeculae region; Step S4, constructing a deep learning model to perform morphological processing on the segmented trabecular bone area and refine the boundary of the trabecular bone area; Step S5, performing multiple iterations of feature extraction on the segmented trabecular bone region, and adjusting feature extraction parameters according to the result of the previous iteration; Step S6, performing correlation analysis on the extracted trabecular bone features, and displaying the distribution and change trend of the features through a data visualization method; Wherein, in step S5, the following sub-steps are also included: S5-1, performing the first iteration feature extraction, extracting the morphological features of trabeculae from the segmented image, wherein the morphological features include trabeculae thickness, trabeculae number and trabeculae separation, and the specific formula is: ; ; ; Among them, TbTh represents the thickness of trabeculae, TbN represents the number of trabeculae, and TbSp represents the separation of trabeculae. is the area of ​​the trabecular bone region, is the total length of the trabeculae, is the area of ​​the entire image, is the area of ​​the background region; S5-2, performing the first iteration feature extraction, extracting the structural features of trabeculae from the segmented image, wherein the structural features include trabeculae connection density and anisotropy, and the specific formula is: ; ; Among them, Xv represents the Euler characteristic of the trabecular structure, which is used to describe the connection density of the trabecular structure, DA represents the anisotropy, is the 0th-order Betti number, which represents the number of connected regions. is the second-order Betti number, which indicates the number of holes. and are the eigenvalues ​​of the inertia tensor of the trabecular structure, is the maximum eigenvalue, indicating the distribution of trabeculae in the main direction, is the second largest eigenvalue, indicating the distribution of trabeculae in the secondary direction; S5-3, performing the first iteration feature extraction, extracting the texture features of trabeculae from the segmented image, wherein the texture features include wavelet transform features and trabeculae scores, and the specific formula is: ; ; ; ; in, are the wavelet coefficients, represents a two-dimensional image, is the value of the wavelet basis function at position (j, k), E is the wavelet energy, H is the wavelet entropy, is the wavelet coefficient The probability distribution of , TBS represents the trabecular bone score, V(e) is the experimental variation function, which represents the grayscale change of the distance e function; S5-4, adjusting feature extraction parameters according to the feature results extracted in the first iteration; S5-5, performing a second iteration of feature extraction, repeating steps S5-1 to S5-3, and extracting trabecular bone features after adjusting feature extraction parameters; S5-6, repeating steps S5-4 and S5-5, performing multiple iterative feature extractions until the standard deviation of the extraction results of the same feature in multiple iterations is less than a preset value, wherein the preset value is 0.005.

[0009] Furthermore, in step S1, the following sub-steps are also included: S1-1, setting scanning parameters of the X-ray imaging device and calibrating the scanning parameters using a standard calibration sample, wherein the scanning parameters include voltage, current and exposure time; S1-2, start the X-ray imaging device to scan the bone tissue and obtain a two-dimensional projection image of the bone tissue; S1-3, storing the scanned image data in DICOM format, storing it in a local server, and backing up the stored image data, wherein the backup is stored in an external storage device.

[0010] Furthermore, in step S2, the following sub-steps are also included: S2-1, setting parameters of a Gaussian filter, using a convolution operation to perform weighted averaging of each pixel in the two-dimensional projection image of the bone tissue and pixels in its neighborhood through the Gaussian filter to reduce noise in the image, wherein the parameters include a standard deviation and a filter window size; S2-2, the grayscale value range of the denoised image is adjusted by an adaptive contrast enhancement method to increase the image contrast and highlight the structural features of trabecular bone; S2-3, performing histogram equalization on the enhanced image, by adjusting the gray value distribution of the image so that the histogram of the image is evenly distributed; S2-4, normalize the image after histogram equalization and adjust the grayscale value range of the image to [0,1].

[0011] Furthermore, in step S3, the following sub-steps are also included: S3-1, calculating the preprocessed image gradient by edge detection algorithm, identifying the boundary of trabeculae, and generating an edge map, wherein the edge detection algorithm includes Canny algorithm and Sobel algorithm; S3-2, performing binarization processing on the image after edge detection, converting the edge image into a binary image by setting a threshold through an adaptive threshold method, and distinguishing the trabecular bone area from the background area; S3-3, traverse each pixel in the binarized image through depth-first search to identify connected trabecular bone regions; S3-4, marking the identified connected areas through a marking algorithm and assigning a unique identifier; S3-5, calculating the area of ​​each connected region, removing isolated noise points and small area quantities whose areas are smaller than a preset threshold, and separating the connected trabecular bone regions.

[0012] Furthermore, in step S4, the following sub-steps are also included: S4-1: Design an encoder path of a deep learning model based on an improved DeepLab v3+ neural network structure for extracting contextual information of an image, wherein the encoder path includes a convolutional layer and a dilated convolutional layer; S4-2: Design a decoder path of a deep learning model based on an improved DeepLab v3+ neural network structure for recovering image information, wherein the decoder path includes an upsampling layer and a convolutional layer; S4-3: Introduce a skip connection between the encoder and the decoder to concatenate the feature map in the encoder path with the feature map in the decoder path; S4-4: Introduce the spatial pyramid pooling ASPP module to capture feature information at different scales; S4-5, performing model training, using the training data set to train the model until the Dice coefficient reaches a preset value, the preset value being 0.9; S4-6, using the trained deep learning model to perform morphological processing on the segmented trabecular bone area to refine the boundary of the trabecular bone area, the morphological processing includes expansion and erosion.

[0013] Furthermore, in step S6, the following sub-steps are also included: S6-1, normalize the extracted trabecular morphological features, structural features, and texture features, and scale the feature values ​​to the range of [0,1]; S6-2, the correlation among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was calculated using the Pearson correlation coefficient, and the relationship among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was evaluated according to the value of the correlation coefficient; S6-3, displaying the morphological features, structural features and texture features through data visualization methods, wherein the data visualization methods include heat maps, scatter plots, histograms, line graphs, bar graphs and box plots.

[0014] The beneficial effects brought about by the technical solution provided by the present invention include at least: The present invention uses X-ray imaging equipment to obtain high-resolution two-dimensional projection images, and combines them with an improved deep learning model for image segmentation and feature extraction. It can more accurately identify and separate trabecular bone areas. Through multiple iterations of feature extraction and parameter adjustment, the stability and accuracy of the feature extraction results are ensured, thereby providing a more reliable basis for the diagnosis of orthopedic diseases such as osteoporosis.

[0015] During the feature extraction process, the present invention not only focuses on the morphological characteristics of trabeculae, but also extracts structural characteristics and texture characteristics, which can more comprehensively evaluate the health status of trabeculae. Through the deep learning model, the segmented trabecular area is morphologically processed, the boundaries of the trabecular area are refined, and the accuracy of feature extraction is further improved.

[0016] By introducing edge detection algorithms and deep learning models, the present invention can more accurately separate trabecular bone areas and effectively identify the boundaries of trabecular bone. Through depth-first search and labeling algorithms, connected trabecular bone areas can be accurately identified and labeled, thereby improving the accuracy of image segmentation.

[0017] The present invention significantly improves the computational efficiency of feature extraction by optimizing the image processing flow and introducing an efficient deep learning model. Combined with adaptive contrast enhancement and histogram equalization methods, the image quality can be quickly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A flow chart of a method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the trabecular feature extraction method based on image analysis proposed by the present invention, its specific implementation method, structure, characteristics and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0021] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0022] The following examples are for illustrative purposes only and are not intended to limit the scope of the present invention.

[0023] The following is a detailed description of a method for extracting trabecular bone features based on image analysis provided by the present invention in conjunction with the accompanying drawings.

[0024] See also Figure 1 , which shows a method flow chart of a method for extracting trabecular features based on image analysis provided by an embodiment of the present invention, the method comprising the following steps: The process of using the standard calibration sample includes placing the standard calibration sample in the scanning device, scanning it according to the set scanning parameters, and adjusting the scanning parameters by analyzing the scanning results of the calibration sample to ensure that the performance of the device reaches the optimal state.

[0025] The use of X-ray imaging equipment includes 1. Equipment startup: Start the X-ray imaging equipment according to the equipment's operating manual to ensure that the equipment is in normal working condition; 2. Place the sample: Place the bone tissue sample to be scanned in the scanning area of ​​the scanning device, and ensure that the position and direction of the sample meet the scanning requirements; 3. Start scanning: Start the scanning program, and the device scans the bone tissue according to the preset parameters (voltage, current and exposure time) to obtain a two-dimensional projection image; 4. Real-time monitoring: During the scanning process, the image quality is monitored in real time to ensure that there is no obvious artifact or noise interference; Storage format of image data: Scanned image data is stored in DICOM format, which is a standard format for storing and transmitting medical images. It has wide compatibility and scalability. Using DICOM format to store image data can ensure the integrity and accuracy of image information. Image data storage: Scanned image data is stored on a local server, which can provide fast data access and processing capabilities, making it easier for researchers and clinicians to view and analyze image data in a timely manner; Image data backup: Back up the image data stored on the local server and store the backup on an external storage device (external hard disk, network attached storage NAS). The backup operation can ensure the security and reliability of the data and prevent data loss due to equipment failure or human error.

[0026] Step S2, preprocessing the acquired two-dimensional projection image of bone tissue, including denoising, contrast enhancement, histogram equalization and normalization processing; Wherein step S2 also includes the following sub-steps: S2-1, setting the parameters of the Gaussian filter, using the convolution operation of the Gaussian filter to perform weighted averaging on each pixel in the two-dimensional projection image of the bone tissue and the pixels in its neighborhood, so as to reduce the noise in the image, and the parameters include the standard deviation and the filter window size; S2-2, the grayscale value range of the denoised image is adjusted by an adaptive contrast enhancement method to increase the image contrast and highlight the structural features of trabecular bone; S2-3, performing histogram equalization on the enhanced image, by adjusting the gray value distribution of the image so that the histogram of the image is evenly distributed; S2-4, normalize the image after histogram equalization and adjust the grayscale value range of the image to [0,1].

[0027] It should be noted that the Gaussian filter is a commonly used image smoothing technology used to reduce noise in images. The denoising effect is achieved by taking a weighted average of each pixel with the pixels in its neighborhood. The weight of the Gaussian filter is determined by the Gaussian function. The weight of the central pixel is the largest, and the weight of the neighborhood pixels decreases with increasing distance.

[0028] The standard deviation determines the degree of smoothing of the Gaussian filter. A larger standard deviation will result in a stronger smoothing effect but blur the details in the image, while a smaller standard deviation will retain more details but the denoising effect will be relatively weaker.

[0029] The filter window size determines the number of pixels in the neighborhood that participate in the weighted average. A larger window can better remove noise, but the computational cost is higher. A smaller window is more efficient, but the denoising effect may not be ideal. The filter window size is an odd number.

[0030] The adaptive contrast enhancement method is a technology for dynamically adjusting the contrast of an image, which can automatically adjust the grayscale value range according to the local features of the image, thereby enhancing the contrast of the image and highlighting the structural features of interest; in the present invention, the adaptive contrast enhancement method is used to adjust the grayscale value range of the denoised two-dimensional projection image of bone tissue to increase the contrast of the image and highlight the structural features of trabeculae.

[0031] Histogram equalization is a commonly used image enhancement technology. By adjusting the gray value distribution of the image to make the histogram of the image evenly distributed, the contrast of the image can be further improved and the details in the image can be made clearer.

[0032] Normalization adjusts the grayscale value range of the image to [0,1]. The unified grayscale value range can reduce the differences between different images and improve the robustness of the algorithm. The grayscale value range after normalization is smaller, which can improve the efficiency of subsequent calculations. Especially when it comes to deep learning models, normalization is a necessary preprocessing step.

[0033] Step S3, segmenting the image using an edge detection segmentation method to separate the trabeculae from the image to obtain a trabeculae region; Wherein step S3 also includes the following sub-steps: S3-1, calculating the preprocessed image gradient by edge detection algorithm, identifying the boundary of trabeculae, and generating edge map, the edge detection algorithm includes Canny algorithm and Sobel algorithm; S3-2, performing binarization processing on the image after edge detection, converting the edge image into a binary image by setting a threshold through an adaptive threshold method, and distinguishing the trabecular bone area from the background area; S3-3, traverse each pixel in the binarized image through depth-first search to identify connected trabecular bone regions; Step S1, scanning the bone tissue using an X-ray imaging device to obtain a two-dimensional projection image of the bone tissue; Wherein step S1 also includes the following sub-steps: S1-1, setting scanning parameters of the X-ray imaging device and calibrating the scanning parameters using a standard calibration sample, the scanning parameters including voltage, current and exposure time; S1-2, start the X-ray imaging device to scan the bone tissue and obtain a two-dimensional projection image of the bone tissue; S1-3, storing the scanned image data in DICOM format in a local server, and backing up the stored image data in an external storage device.

[0034] It should be noted that voltage affects the penetration ability of X-rays. Higher voltage can increase the penetration ability of X-rays and is suitable for thicker or denser bone tissue, but too high a voltage may lead to a decrease in image contrast. The voltage range is between 50-150 kV, and the specific value needs to be adjusted according to the density and thickness of the bone tissue.

[0035] The current determines the intensity of the X-rays. A higher current can improve the signal-to-noise ratio of the image, but it will also increase the radiation dose. Under the premise of ensuring image quality, a lower current value should be selected as much as possible, and the current range should be between 10-100 mA.

[0036] Exposure time affects the brightness and contrast of the image. A longer exposure time can increase the brightness of the image, but it will also increase noise. It is necessary to select an appropriate exposure time based on the density of bone tissue and the required image quality, usually between 10-100ms.

[0037] Calibration samples: Calibration is performed using standard calibration samples, which have known density and structural characteristics. Through calibration, the scanning parameters can be adjusted to ensure the performance of the equipment is optimal.

[0038] S3-4, marking the identified connected areas through a marking algorithm and assigning a unique identifier; S3-5, calculating the area of ​​each connected region, removing isolated noise points and small area quantities whose areas are smaller than a preset threshold, and separating the connected trabecular bone regions.

[0039] It should be noted that edge detection is an important step in image processing, which is used to identify areas in the image where the grayscale value changes significantly. These areas correspond to the boundaries of objects. In trabecular bone images, edge detection can help identify the boundaries between trabeculae and the background.

[0040] Edge detection algorithms identify boundaries by calculating the gradient of an image and generate an edge map, which is a binary image in which edge pixels are marked as 1 and non-edge pixels are marked as 0.

[0041] The Canny algorithm is a classic edge detection algorithm that detects edges through multiple steps, including Gaussian filtering denoising, calculating gradient amplitude and direction, non-maximum suppression, and dual threshold detection. It can detect clear edges and has a good noise suppression effect.

[0042] The Sobel algorithm is an edge detection algorithm based on gradient calculation. It detects edges by calculating the gradients of an image in the horizontal and vertical directions. The Sobel operator is sensitive to horizontal and vertical edges in an image, has simple calculations, and is suitable for fast edge detection.

[0043] Binarization is the process of converting an image into a binary image, in which the pixel values ​​are only 0 and 1. This can simplify the image processing process and facilitate the subsequent identification and marking of connected regions.

[0044] The adaptive threshold method can dynamically set the threshold according to the local characteristics of the image instead of using a global fixed threshold, which can better adapt to the illumination changes and noise effects in the image.

[0045] Distinguish trabecular bone area from background area: By setting a threshold, the trabecular bone area can be distinguished from the background area.

[0046] Depth-first search (DFS) is a traversal algorithm used to identify connected regions in binary images. In trabecular bone images, DFS can identify connected regions of trabecular bone, providing a basis for subsequent labeling and analysis.

[0047] Depth-first search consists of the following steps: 1. Initialization: Starting from the upper left corner pixel of the binary image, traverse the image pixel by pixel; 2. Search process: For each unvisited pixel, if its value is 1 (i.e., it belongs to the trabecular bone area), a depth-first search is performed starting from that pixel to mark all connected pixels; 3. Recursive search: For the current pixel, check its 8 neighboring pixels (upper, lower, left, right, upper left, lower left, upper right, lower right). If the neighboring pixel value is 1 and has not been visited, add the neighboring pixel to the search queue and continue searching its neighboring pixels until all connected pixels are visited; A labeling algorithm is an algorithm used to identify and label connected regions in an image. It divides the pixels in the image into different regions by assigning a unique identifier (label) to each connected region.

[0048] The marking algorithm consists of the following steps: 1. Labeling process: During the depth-first search process, a unique identifier is assigned to each newly discovered connected region, and all pixels in the connected region are labeled with the identifier; 2. Label matrix: Create a label matrix of the same size as the binary image to store the connected region identifier to which each pixel belongs; Area calculation: Calculate the area of ​​each connected region, that is, the number of pixels in the region, which can be achieved by counting the number of times each identifier appears in the label matrix; Noise removal: Remove isolated noise points and small areas with an area smaller than a preset threshold. These areas may be caused by noise or inaccurate image segmentation. The preset threshold can be determined based on the minimum size of trabeculae.

[0049] Step S4, constructing a deep learning model to perform morphological processing on the segmented trabecular bone area and refine the boundary of the trabecular bone area; Wherein step S4 also includes the following sub-steps: S4-1: Design the encoder path of the deep learning model based on the improved DeepLab v3+ neural network structure to extract the context information of the image. The encoder path includes convolutional layers and dilated convolutional layers. S4-2: Design a decoder path of a deep learning model based on an improved DeepLab v3+ neural network structure to recover image information. The decoder path includes upsampling layers and convolutional layers. S4-3: Introduce a skip connection between the encoder and the decoder to concatenate the feature map in the encoder path with the feature map in the decoder path; S4-4: Introduce the spatial pyramid pooling ASPP module to capture feature information at different scales; S4-5, performing model training, using the training data set to train the model until the Dice coefficient reaches a preset value, the preset value being 0.9; S4-6, using the trained deep learning model to perform morphological processing on the segmented trabecular bone area to refine the boundary of the trabecular bone area, the morphological processing includes expansion and erosion.

[0050] It should be noted that the main function of the encoder path is to extract rich contextual information from the input image. Through multi-layer convolution operations, the model can learn the local and global features of the image, providing a basis for subsequent segmentation tasks.

[0051] Convolution layer: extracts local features by sliding the convolution kernel on the input feature map. The convolution operation can capture the edge and texture information in the image. Dilated convolution layer: Dilated convolution is a special convolution operation that introduces holes in the convolution kernel (i.e. skipping some pixels) to expand the receptive field of the convolution kernel, thereby capturing a wider range of contextual information without increasing the amount of computation. Dilated convolution is particularly suitable for tasks that require a large receptive field. Restoring image information: The main function of the decoder path is to gradually restore the high-level features extracted by the encoder path to the resolution of the original image. Through upsampling operations, the model can generate segmentation results with the same resolution as the input image. Upsampling layer: used to increase the spatial resolution of the feature map. Upsampling methods include nearest neighbor interpolation, bilinear interpolation, and transposed convolution; Convolutional layer: In the decoder path, the convolutional layer is used to refine the upsampled feature map and remove artifacts that may be introduced during the upsampling process, making the segmentation result more accurate.

[0052] Skip connections are a method of passing low-level features from the encoder path directly into the decoder path. Through skip connections, the model is able to retain more detail information, thereby better recovering details when restoring image resolution.

[0053] Skip connections are able to combine detailed features in the encoder path with high-level features in the decoder path, thereby improving the segmentation accuracy, especially for the segmentation of small objects and complex structures.

[0054] The Spatial Pyramid Pooling (ASPP) module is a structure used to capture multi-scale features. It extracts feature information of different scales through parallel atrous convolutional layers and pooling layers. The ASPP module can effectively process targets of different scales and improve the robustness of the model.

[0055] During training, the model optimizes parameters by minimizing the difference between the predicted segmentation results and the true annotations (using Dice loss) and is trained using the Adam optimization algorithm.

[0056] The training data set is a two-dimensional projection image of the labeled trabecular bone area, which is used to provide a data set for training the deep learning model to ensure that the model can learn the characteristics of the trabecular bone area.

[0057] The Dice coefficient is a commonly used indicator to measure segmentation accuracy, which calculates the similarity between the predicted segmentation result and the true annotation.

[0058] Morphological processing is a processing method based on image shape. Through dilation and corrosion operations, the boundary of the trabecular area can be refined to make the segmentation result more accurate.

[0059] The erosion operation is an image processing method based on the structural element. It is used to remove small objects and boundary pixels in the image. The erosion effect is achieved by sliding the structural element on the image and taking the minimum value of all pixel values ​​covered by the structural element.

[0060] The dilation operation is the opposite of the erosion operation and is used to fill small holes and broken connections in an image. The dilation effect is achieved by sliding the structural element across the image and taking the maximum value of all pixels covered by the structural element.

[0061] Step S5, performing multiple iterations of feature extraction on the segmented trabecular bone region, and adjusting feature extraction parameters according to the result of the previous iteration; Wherein, in step S5, the following sub-steps are also included: S5-1, perform the first iteration feature extraction, extract the morphological features of trabeculae from the segmented image, the morphological features include trabeculae thickness, trabeculae number and trabeculae separation, the specific formula is: ; ; ; Among them, TbTh represents the thickness of trabeculae, TbN represents the number of trabeculae, and TbSp represents the separation of trabeculae. is the area of ​​the trabecular bone region, is the total length of the trabeculae, is the area of ​​the entire image, is the area of ​​the background region; S5-2, perform the first iteration feature extraction, extract the structural features of trabeculae from the segmented image, the structural features include trabecular connection density and anisotropy, the specific formula is: ; ; Among them, Xv represents the Euler characteristic of the trabecular structure, which is used to describe the connection density of the trabecular structure, DA represents the anisotropy, is the 0th-order Betti number, which represents the number of connected regions. is the second-order Betti number, which indicates the number of holes. and are the eigenvalues ​​of the inertia tensor of the trabecular structure, is the maximum eigenvalue, indicating the distribution of trabeculae in the main direction, is the second largest eigenvalue, indicating the distribution of trabeculae in the secondary direction; S5-3, perform the first iteration feature extraction, extract the texture features of trabeculae from the segmented image, the texture features include wavelet transform features and trabeculae score, the specific formula is: ; ; ; ; in, are the wavelet coefficients, represents a two-dimensional image, is the value of the wavelet basis function at position (j, k), E is the wavelet energy, H is the wavelet entropy, is the wavelet coefficient The probability distribution of , TBS represents the trabecular bone score, V(e) is the experimental variation function, which represents the grayscale change of the distance e function; S5-4, adjusting feature extraction parameters according to the feature results extracted in the first iteration; S5-5, performing a second iteration of feature extraction, repeating steps S5-1 to S5-3, and extracting trabecular bone features after adjusting feature extraction parameters; S5-6, repeating steps S5-4 and S5-5, performing multiple iterative feature extractions until the standard deviation of the extraction results of the same feature in multiple iterations is less than a preset value, the preset value being 0.005.

[0062] It should be noted that trabecular thickness refers to the average thickness of trabeculae, which reflects the thickness of trabeculae. The thickness of trabeculae is determined by measuring the average width of the trabecular area.

[0063] Trabecular number: refers to the number of trabeculae per unit area, reflecting the density of trabeculae. The density of trabeculae is determined by counting the number of trabeculae per unit area. Trabecular separation: It indicates the average spacing between trabeculae, reflecting the density of trabeculae distribution. The degree of trabecular separation is determined by measuring the average spacing between trabeculae. Trabecular connection density: It indicates the connection density of trabecular structure, reflecting the degree of connection between trabeculae. The Euler characteristic is obtained by calculating the 0th and second-order Betti numbers of trabecular structure. Anisotropy: It indicates the degree of anisotropy of the trabecular structure and reflects the arrangement direction of the trabecular structure. The inertia tensor of the trabecular structure is calculated to obtain its eigenvalue. The anisotropy is calculated based on the eigenvalue of the inertia tensor. Wavelet transform features: The texture information of the image is extracted through wavelet transform to reflect the texture details of the trabeculae. The segmented trabeculae area is subjected to wavelet transform to extract the wavelet coefficients. The wavelet energy is calculated through the sum of the squares of the wavelet coefficients. The wavelet entropy is calculated through the probability distribution of the wavelet coefficients. Trabecular bone score: The texture uniformity of trabecular bone is evaluated by experimental variogram, which reflects the texture quality of trabecular bone. The trabecular bone score is calculated by experimental variogram; Adjust parameters: For morphological feature extraction, the threshold parameters of image segmentation can be adjusted, including the global threshold and adaptive threshold, to better separate the trabecular area and the background area; the morphological operation parameters, including the size of the structural element and the number of operations, can be adjusted to better refine the trabecular boundary; the feature calculation parameters, including the step size of the trabecular thickness calculation, the grid resolution of the trabecular number statistics, and the neighborhood range of the trabecular separation calculation can be adjusted to improve the precision and accuracy of the feature calculation.

[0064] For structural feature extraction, the neighborhood size and connection weight for connection density calculation can be adjusted to more accurately reflect the degree of connection between trabeculae; the anisotropy calculation parameters can be adjusted to more accurately reflect the arrangement direction of trabeculae; and the Euler feature calculation parameters can be adjusted to more accurately describe the connection density of trabecular structure.

[0065] For texture feature extraction, the scale parameter of wavelet transform, wavelet basis function and direction parameter of wavelet transform can be adjusted to better capture the texture characteristics of trabecular bone; the texture feature calculation parameters, including the weight of wavelet energy calculation, the normalization method of wavelet entropy calculation and the calculation method of texture score can be adjusted to more accurately evaluate the texture quality of trabecular bone.

[0066] Compare results: Compare the feature results extracted in the second iteration with those of the first iteration to evaluate the effect of parameter adjustment and compare the standard deviation of the feature values ​​to check if there is any improvement.

[0067] Calculate standard deviation: After each iteration, calculate the standard deviation of the extraction results of the same feature in multiple iterations, and check whether the calculated standard deviation is less than the preset value of 0.005. If it is satisfied, stop the iteration; if not, continue to the next iteration.

[0068] Step S6, performing correlation analysis on the extracted trabecular bone features, and displaying the distribution and change trend of the features through a data visualization method; Wherein, in step S6, the following sub-steps are also included: S6-1, normalize the extracted trabecular morphological features, structural features, and texture features, and scale the feature values ​​to the range of [0,1]; S6-2, the correlation among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was calculated using the Pearson correlation coefficient, and the relationship among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was evaluated according to the value of the correlation coefficient; S6-3, the morphological features, structural features and texture features are displayed through data visualization methods, including heat maps, scatter plots, histograms, line graphs, bar graphs and box plots.

[0069] It should be noted that normalization scales all eigenvalues ​​to the same range ([0,1]). Different features may have different dimensions and ranges. Normalization can eliminate such differences and improve the efficiency of subsequent calculations.

[0070] The Pearson correlation coefficient is used to evaluate the linear correlation between two features. Its value range is between [-1, 1], where 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation.

[0071] Data visualization can intuitively display the distribution and changing trends of features, helping researchers and clinicians better understand the data. Through visualization, the relationship between features, outliers, and the laws of data distribution can be quickly identified.

[0072] Heat map: The heat map shows the correlation matrix between features, which intuitively indicates the strength of the correlation between features. Scatter plot: The relationship between two features can be shown through a scatter plot, and the distribution and correlation between the features can be observed intuitively; Histogram: The distribution of a single feature is displayed through a histogram, and the frequency distribution of the feature can be observed intuitively; Line chart: A line chart is used to show the changing trend of features over time or samples. It is suitable for displaying dynamic data. Bar chart: The bar chart shows the difference in feature values ​​between different categories or groups, which is suitable for comparative analysis; Box plot: The box plot shows the statistical distribution of features, including the median, quartiles, and outliers. It is suitable for showing the distribution of data.

[0073] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for extracting trabecular bone features based on image analysis, characterized in that: The method includes: Step S1, scanning the bone tissue using an X-ray imaging device to obtain a two-dimensional projection image of the bone tissue; Step S2, preprocessing the acquired two-dimensional projection image of bone tissue, including denoising, contrast enhancement, histogram equalization and normalization processing; Step S3, segmenting the image using an edge detection segmentation method to separate the trabeculae from the image to obtain a trabeculae region; Step S4, constructing a deep learning model to perform morphological processing on the segmented trabecular bone area and refine the boundary of the trabecular bone area; Step S5, performing multiple iterations of feature extraction on the segmented trabecular bone region, and adjusting feature extraction parameters according to the result of the previous iteration; Step S6, performing correlation analysis on the extracted trabecular bone features, and displaying the distribution and change trend of the features through a data visualization method; Wherein, in step S5, the following sub-steps are also included: S5-1, performing the first iteration feature extraction, extracting the morphological features of trabeculae from the segmented image, wherein the morphological features include trabeculae thickness, trabeculae number and trabeculae separation, and the specific formula is: ； ； ； Among them, TbTh represents the thickness of trabeculae, TbN represents the number of trabeculae, and TbSp represents the separation of trabeculae. is the area of ​​the trabecular bone region, is the total length of the trabeculae, is the area of ​​the entire image, is the area of ​​the background region; S5-2, performing the first iteration feature extraction, extracting the structural features of trabeculae from the segmented image, wherein the structural features include trabeculae connection density and anisotropy, and the specific formula is: ； ； Among them, Xv represents the Euler characteristic of the trabecular structure, which is used to describe the connection density of the trabecular structure, DA represents the anisotropy, is the 0th-order Betti number, which represents the number of connected regions. is the second-order Betti number, indicating the number of holes, and are the eigenvalues ​​of the inertia tensor of the trabecular structure, is the maximum eigenvalue, indicating the distribution of trabeculae in the main direction, is the second largest eigenvalue, indicating the distribution of trabeculae in the secondary direction; S5-3, performing the first iteration feature extraction, extracting the texture features of trabeculae from the segmented image, wherein the texture features include wavelet transform features and trabeculae scores, and the specific formula is: ； ； ； ； in, are the wavelet coefficients, represents a two-dimensional image, is the value of the wavelet basis function at position (j, k), E is the wavelet energy, H is the wavelet entropy, is the wavelet coefficient The probability distribution of , TBS represents the trabecular bone score, V(e) is the experimental variation function, which represents the grayscale change of the distance e function; S5-4, adjusting feature extraction parameters according to the feature results extracted in the first iteration; S5-5, performing a second iteration of feature extraction, repeating steps S5-1 to S5-3, and extracting trabecular bone features after adjusting feature extraction parameters; S5-6, repeating steps S5-4 and S5-5, performing multiple iterative feature extractions until the standard deviation of the extraction results of the same feature in multiple iterations is less than a preset value, wherein the preset value is 0.

005.

2. The method for extracting trabecular bone features based on image analysis according to claim 1, characterized in that: Wherein step S1 also includes the following sub-steps: S1-1, setting scanning parameters of the X-ray imaging device and calibrating the scanning parameters using a standard calibration sample, wherein the scanning parameters include voltage, current and exposure time; S1-2, start the X-ray imaging device to scan the bone tissue and obtain a two-dimensional projection image of the bone tissue; S1-3, storing the scanned image data in DICOM format, storing it in a local server, and backing up the stored image data, wherein the backup is stored in an external storage device.

3. The method for extracting trabecular bone features based on image analysis according to claim 1, characterized in that: Wherein step S2 also includes the following sub-steps: S2-1, setting parameters of a Gaussian filter, using a convolution operation to perform weighted averaging of each pixel in the two-dimensional projection image of the bone tissue and pixels in its neighborhood through the Gaussian filter to reduce noise in the image, wherein the parameters include a standard deviation and a filter window size; S2-2, the grayscale value range of the denoised image is adjusted by an adaptive contrast enhancement method to increase the image contrast and highlight the structural features of trabecular bone; S2-3, performing histogram equalization on the enhanced image, by adjusting the gray value distribution of the image so that the histogram of the image is evenly distributed; S2-4, normalize the image after histogram equalization and adjust the grayscale value range of the image to [0,1].

4. The method for extracting trabecular bone features based on image analysis according to claim 1, characterized in that: Wherein step S3 also includes the following sub-steps: S3-1, calculating the preprocessed image gradient by edge detection algorithm, identifying the boundary of trabeculae, and generating an edge map, wherein the edge detection algorithm includes Canny algorithm and Sobel algorithm; S3-2, performing binarization processing on the image after edge detection, converting the edge image into a binary image by setting a threshold value through an adaptive threshold method, and distinguishing the trabecular bone area from the background area; S3-3, traverse each pixel in the binarized image through depth-first search to identify connected trabecular bone regions; S3-4, marking the identified connected areas through a marking algorithm and assigning a unique identifier; S3-5, calculating the area of ​​each connected region, removing isolated noise points and small area quantities whose areas are smaller than a preset threshold, and separating the connected trabecular bone regions.

5. The method for extracting trabecular bone features based on image analysis according to claim 1, characterized in that: Wherein step S4 also includes the following sub-steps: S4-1: Design an encoder path of a deep learning model based on an improved DeepLab v3+ neural network structure for extracting contextual information of an image, wherein the encoder path includes a convolutional layer and a dilated convolutional layer; S4-2: Design a decoder path of a deep learning model based on an improved DeepLab v3+ neural network structure for recovering image information, wherein the decoder path includes an upsampling layer and a convolutional layer; S4-3: Introduce a skip connection between the encoder and the decoder to concatenate the feature map in the encoder path with the feature map in the decoder path; S4-4: Introduce the spatial pyramid pooling ASPP module to capture feature information of different scales; S4-5, performing model training, using the training data set to train the model until the Dice coefficient reaches a preset value, the preset value being 0.9; S4-6, using the trained deep learning model to perform morphological processing on the segmented trabecular bone area to refine the boundary of the trabecular bone area, the morphological processing includes expansion and erosion.

6. The method for extracting trabecular bone features based on image analysis according to claim 1, characterized in that: Wherein, in step S6, the following sub-steps are also included: S6-1, normalize the extracted trabecular morphological features, structural features, and texture features, and scale the feature values ​​to the range of [0,1]; S6-2, the correlation among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was calculated using the Pearson correlation coefficient, and the relationship among trabecular bone morphological characteristics, structural characteristics, and texture characteristics was evaluated according to the value of the correlation coefficient; S6-3, displaying the morphological features, structural features and texture features through data visualization methods, wherein the data visualization methods include heat maps, scatter plots, histograms, line graphs, bar graphs and box plots.

Images