Bone image enhancement method and system
Through computed tomography and neural network technology, only bone voxels are retained and non-skeletal noise is eliminated, which solves the problem of low X-ray image clarity and improves the clarity and accuracy of bone images at the software level.
Patent Information
- Application Number
- CN202411198811.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The sharpness of X-ray images is low, making it difficult to meet the accuracy requirements of modern medical care, especially for patients with different thicknesses and old equipment.
CT volume data was obtained through computed tomography, and simulated X-ray imaging was performed after only bone voxel was retained. The neural network was trained to obtain bone image enhancement, eliminate non-skeletal tissue noise, and register the original X-ray image to obtain enhanced bone structure images.
Without increasing the patient's radiation dose, clear bone X-ray images are provided to enhance bone tissue structure and boundaries, and maintain the development of other tissues. Only software processing is required without hardware updates.
Smart Images

Figure CN119027406A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of medical imaging, and in particular relates to a bone imaging enhancement method and system. Background Art
[0002] X-ray examination is a common imaging auxiliary examination method in clinical practice. Due to the special imaging principle of X-ray, the clarity of X-ray images depends on the voltage, current, exposure time and other parameters set by the X-ray equipment during shooting. In order to obtain clearer X-ray examination images, higher voltage, current and longer exposure time are often set to increase the power of the X-ray examination equipment, but higher power will also greatly increase the radiation dose received by the patient.
[0003] In addition, since X-ray imaging is performed by emitting X-rays that penetrate human tissues, different patients are fat or thin, and the thickness of human tissues varies. When X-rays pass through human tissues of different thicknesses, they are absorbed to different degrees, and the clarity of the images is also different. Thinner patients will have clearer images, while fatter patients often cannot obtain clear enough X-ray images because the default power of the X-ray equipment is insufficient.
[0004] At the same time, most of the lower-level medical institutions in the country have outdated equipment. Due to the power limitations of the equipment itself and outdated technology, the clarity of the X-ray images taken can no longer meet the precision requirements of modern medicine. Summary of the invention
[0005] The technical purpose of the present invention is to provide a bone image enhancement method and system to solve the problem that the clarity of X-ray images is low and it is difficult to meet the accuracy requirements of modern medical treatment.
[0006] To solve the above problems, the technical solution of the present invention is:
[0007] A bone image enhancement method comprises the following steps:
[0008] Acquire a number of CT volume data by computer tomography, perform data processing on each CT volume data, and obtain processed CT volume data retaining only bone voxels; process the CT volume data and the processed CT volume data respectively by simulating X-ray imaging, and obtain a plurality of groups of first pseudo X-ray images and second pseudo X-ray images respectively;
[0009] Using multiple groups of first pseudo X-ray images and second pseudo X-ray images as training sets, and finding a mapping relationship from the first pseudo X-ray image to the second pseudo X-ray image as a training goal, a neural network with bone image enhancement is obtained by training;
[0010] The original X-ray image to be enhanced is input into the trained neural network, and the corresponding bone-enhanced X-ray image is output;
[0011] The bone-enhanced X-ray image is registered to the original X-ray image to obtain an X-ray image of the bone structure with enhanced development.
[0012] The data processing of the CT volume data specifically includes: selecting non-bone tissue in the CT volume data, and setting the voxel value of the non-bone tissue to the lowest HU value, thereby obtaining processed CT volume data that only retains the bone voxels.
[0013] In the process of training the first pseudo X-ray image and the second pseudo X-ray image, random noise is also input, and the algorithm simulates the process of removing redundant tissue images from CT volume data to obtain processed CT volume data, and retains the mapping relationship between the two, thereby obtaining a neural network with bone image enhancement.
[0014] Specifically, the original X-ray image is processed by a neural network to eliminate excess non-bone tissue as noise, wherein the excess non-bone tissue includes soft tissue structure and artifacts.
[0015] Among them, registering the bone-enhanced X-ray image to the original X-ray image specifically includes: selecting the location of the bone structure to be enhanced from the original X-ray image, and finding the corresponding bone structure from the bone-enhanced X-ray image, and then fusing the original X-ray image and the bone-enhanced X-ray image with each other by weighted summation to obtain an X-ray image of the bone structure with enhanced development.
[0016] A bone image enhancement system adapted to the above-mentioned bone image enhancement method comprises: a data set preprocessing module, a bone image enhancement algorithm module, a bone contour enhancement extraction module and an image registration module;
[0017] The data set preprocessing module is configured to obtain CT volume data of the target to be measured obtained from a computer tomography scan, perform data processing on the CT volume data to obtain processed CT volume data retaining only bone voxels; process the CT volume data and the processed CT volume data respectively by simulating X-ray imaging to obtain a series of first pseudo X-ray images and a series of second pseudo X-ray images respectively;
[0018] The bone image enhancement algorithm module is configured to use the first pseudo X-ray image and the second pseudo X-ray image as a training set, to find a mapping relationship from the first pseudo X-ray image to the second pseudo X-ray image as a training goal, and to train a neural network capable of bone image enhancement;
[0019] The bone contour enhancement extraction module is configured to input the original X-ray image to be enhanced into the trained neural network and output the corresponding bone enhanced X-ray image;
[0020] The image registration module is configured to register the bone-enhanced X-ray image to the original X-ray image to obtain an X-ray image of the bone structure with enhanced development.
[0021] A computer device includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the bone image enhancement method as described above.
[0022] A storage medium storing computer-readable instructions, when the computer-readable instructions are executed by one or more processors, enables the one or more processors to execute the bone image enhancement method as described above.
[0023] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:
[0024] The present invention can enhance bone imaging on the basis of the original X-ray film; eliminate multiplication noise and sampling artifacts in the acquisition process through an algorithm to provide clear bone X-ray images; the final imaging only enhances the specific bone tissue structure and boundaries, and does not interfere with the imaging of other tissues in the original image; only conventional X-ray irradiation is required, and bone-enhanced X-ray images can be obtained by processing the original image, without any hardware equipment requirements; X-ray films of any part and any body position (such as anteroposterior, lateral, antero-oblique and bi-oblique positions, etc.) can be enhanced. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.
[0026] Figure 1 A flowchart of a bone image enhancement method of the present invention;
[0027] Figure 2 The flowchart of the bone image enhancement for the spine of the present invention is as follows;
[0028] Figure 3 It is a schematic diagram of the bone image enhancement of the spine according to the present invention. DETAILED DESCRIPTION
[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.
[0030] In order to simplify the drawings, only the parts related to the present invention are schematically shown in each figure, and they do not represent the actual structure of the product. In addition, in order to simplify the drawings and facilitate understanding, in some figures, only one of the parts with the same structure or function is schematically drawn or marked. In this article, "one" not only means "only one", but also means "more than one".
[0031] The following is a further detailed description of a bone image enhancement method and system proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims.
[0032] Example 1
[0033] See also Figure 1 This embodiment provides a method for bone image enhancement, the purpose of which is to obtain relatively clear medical images without increasing the radiation dose received by the patient. In addition to updating the hardware level of the shooting equipment, the image can also be post-processed at the software level to convert low-definition medical images into high-definition images, which specifically includes the following steps:
[0034] First, the CT volume data of the target to be measured is obtained by computer tomography (for subsequent training, a large amount of CT volume data needs to be obtained), and then the CT volume data is processed to obtain processed CT volume data that only retains bone voxels, specifically: non-bone tissue in the CT volume data is selected, and the voxel value of the non-bone tissue is set to the lowest HU value, thereby obtaining processed CT volume data that only retains bone voxels. The principle is: the HU value in CT represents the density difference of different tissues or organs, and the HU value of bones is generally above 400. Therefore, only the part of the CT volume data that is lower than the bone HU value needs to be set to the lowest HU value, so that it will not be imaged, and only the bone voxels will be retained. The CT volume data and the processed CT volume data are projected and processed by simulating X-ray imaging, and a series of first pseudo X-ray images and a series of second pseudo X-ray images are obtained respectively.
[0035] Then, the two types of images, the first pseudo X-ray image and the second pseudo X-ray image in the above steps, are used as training sets and input into an untrained neural network, and random noise is also input. Among them, the neural network can adopt a deep learning network CNN, and all kinds of Encoder-Decoder network structures are possible, such as UNet, nnUNet, etc.; the backbone network can also be a Vision Transformer, such as TransUNet, TransFuse, U-Transformer, etc. Finally, according to the enhanced effect, the standard diffusion UNet structure is adopted in this embodiment. The neural network takes finding the mapping relationship from the first pseudo X-ray image to the second pseudo X-ray image as the training goal, and the algorithm simulates the process of removing redundant tissue images from CT volume data to obtain processed CT volume data. During the training process, the mapping relationship from ordinary X-ray images to bone imaging images only is retained, and a neural network with bone image enhancement function is obtained through training.
[0036] Furthermore, the original X-ray image to be enhanced is input into the trained neural network, the overlapping soft tissue images are removed, and then the corresponding bone-enhanced X-ray image is output. The goal of creating a bone-enhanced X-ray image is to extract a bone model image of a given original X-ray image. To this end, the original X-ray image is processed by the trained neural network, and the redundant non-bone tissue is removed as noise, so that an image that only retains the bone can be obtained, and all soft tissue structures and artifacts will be removed.
[0037] Finally, the bone enhancement X-ray image is registered to the original X-ray image to obtain an X-ray image of the bone structure with enhanced development. Specifically, the location of the bone structure to be enhanced is selected from the original X-ray image, and the corresponding bone structure is found from the bone enhancement X-ray image. Then, the original X-ray image and the bone enhancement X-ray image are fused with each other by weighted summation, so that the original X-ray image allows the bone structure to be selectively enhanced without changing the development effect of other tissues, and finally an X-ray image of the bone structure with enhanced development is obtained. Since the bone enhancement X-ray image does not change the size, resolution, and position of each pixel of the original image, registration only requires the two images to overlap.
[0038] See also Figure 2 and Figure 3, as an example of using this embodiment to enhance the spinal bones in chest X-ray images, first obtain a set of digitally reconstructed chest images for training (including chest CT volume data), model the X-ray imaging process, and obtain a 2D X-ray image (i.e., the first pseudo X-ray image). Then set voxel values for the chest CT volume data to remove redundant tissue images, and then obtain a spinal bone structure image (i.e., the second pseudo X-ray image) by simulating X-ray imaging. Then input the above two images into the neural network for training, and obtain an algorithm model with enhanced spinal bone images. At this time, we only need to send the original X-ray image into the algorithm model for data processing to obtain the spinal bone structure image, and fuse the spinal bone structure image with the original X-ray image to obtain an X-ray image of enhanced spinal bones.
[0039] Example 2
[0040] This embodiment provides a bone image enhancement system, adapted to the bone image enhancement method of embodiment 1, including: a data set preprocessing module, a bone image enhancement algorithm module, a bone contour enhancement extraction module and an image registration module;
[0041] The data set preprocessing module is configured to obtain the CT volume data of the target to be measured obtained by computer tomography, and to generate a 2D first pseudo X-ray image (digital reconstructed image DRR) by simulating the X-ray imaging principle of the volume data for projection processing. Secondly, it is also used to process the CT volume data to obtain processed CT volume data that only retains bone voxels, and then process the processed CT volume data by simulating X-ray imaging to obtain the corresponding second pseudo X-ray image (digital reconstructed image DRR) that only retains the bone image. Since it is necessary to obtain multiple CT volume data, the generated first pseudo X-ray image and second pseudo X-ray image both contain multiple images.
[0042] The bone image enhancement algorithm module is configured to use the first pseudo X-ray image and the second pseudo X-ray image as training sets, along with random noise as input, and the algorithm simulates the process of removing redundant tissue images from CT volume data, retaining the mapping relationship from ordinary X-ray images to bone imaging images only (i.e., the mapping relationship between the first pseudo X-ray image and the second pseudo X-ray image), and finally training a neural network with bone image enhancement.
[0043] The bone contour enhancement extraction module is configured to input the original X-ray image to be enhanced into the trained neural network, remove the overlapping soft tissue images, and obtain the corresponding bone enhanced X-ray image. Here, the goal of constructing a bone enhanced X-ray image is to extract a bone model image of a given original X-ray image. To this end, the original X-ray image is processed by the trained neural network described in the above steps, and the redundant non-bone tissue is eliminated as noise, so that an image that only retains the bones can be obtained, and all soft tissue structures and artifacts are removed from this image.
[0044] The image registration module is configured to register the bone-enhanced X-ray image to the original X-ray image, and the enhanced bone structure processed by the bone contour enhancement extraction module can be added to the original X-ray image, so that the original X-ray image allows the bone structure to be selectively enhanced without changing the development effect of the remaining tissues. Therefore, in order to enhance a real X-ray film, the two images can be fused by using weighted summation to obtain an X-ray image with enhanced bone structure.
[0045] Example 3
[0046] This embodiment provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the bone image enhancement method as described in Embodiment 1.
[0047] Example 4
[0048] A storage medium storing computer-readable instructions, when the computer-readable instructions are executed by one or more processors, causes the one or more processors to execute the bone image enhancement method as described in Example 1.
[0049] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the protection scope of the present invention.
Claims
1. A method for enhancing bone images, characterized in that: The steps include: Acquire a plurality of CT volume data by computer tomography, perform data processing on each of the CT volume data to obtain processed CT volume data retaining only bone voxels; process the CT volume data and the processed CT volume data respectively by simulating X-ray imaging to obtain a plurality of groups of first pseudo X-ray images and second pseudo X-ray images respectively; Using multiple groups of the first pseudo X-ray images and the second pseudo X-ray images as training sets, and finding a mapping relationship from the first pseudo X-ray images to the second pseudo X-ray images as a training goal, to train a neural network capable of bone image enhancement; Inputting the original X-ray image to be enhanced into the trained neural network, and outputting the corresponding bone-enhanced X-ray image; The bone enhancement X-ray image is registered to the original X-ray image to obtain an X-ray image of the bone structure with enhanced development.
2. The bone image enhancement method according to claim 1, characterized in that: The data processing of the CT volume data specifically includes: selecting non-bone tissue in the CT volume data, and setting the voxel value of the non-bone tissue to the lowest HU value, thereby obtaining the processed CT volume data that only retains the bone voxels.
3. The bone image enhancement method according to claim 1, characterized in that: During the training process of the first pseudo X-ray image and the second pseudo X-ray image, random noise is also input, and the algorithm simulates the process of removing redundant tissue images from the CT volume data to obtain the processed CT volume data, and retains the mapping relationship between the two, thereby obtaining the neural network with bone image enhancement.
4. The bone image enhancement method according to claim 1, characterized in that: The original X-ray image is processed by the neural network to eliminate redundant non-bone tissue as noise, wherein the redundant non-bone tissue includes soft tissue structure and artifacts.
5. The bone image enhancement method according to claim 1, characterized in that: The registering of the bone-enhanced X-ray image to the original X-ray image specifically includes: selecting the location of the bone structure to be enhanced from the original X-ray image, finding the corresponding bone structure from the bone-enhanced X-ray image, and then fusing the original X-ray image and the bone-enhanced X-ray image with each other by weighted summation to obtain an X-ray image of the bone structure with enhanced development.
6. A bone image enhancement system, adapted to the bone image enhancement method according to any one of claims 1 to 5, characterized in that: include: Dataset preprocessing module, bone image enhancement algorithm module, bone contour enhancement extraction module and image registration module; The data set preprocessing module is configured to obtain CT volume data of the target to be measured obtained from a computer tomography scan, perform data processing on the CT volume data to obtain processed CT volume data retaining only bone voxels; and process the CT volume data and the processed CT volume data respectively by simulating X-ray imaging to obtain a series of first pseudo X-ray images and a series of second pseudo X-ray images respectively. The bone image enhancement algorithm module is configured to use the first pseudo X-ray image and the second pseudo X-ray image as a training set, to find a mapping relationship from the first pseudo X-ray image to the second pseudo X-ray image as a training goal, and to train a neural network capable of bone image enhancement; The bone contour enhancement extraction module is configured to input the original X-ray image to be enhanced into the trained neural network and output the corresponding bone enhancement X-ray image; The image registration module is configured to register the bone-enhanced X-ray image to the original X-ray image to obtain an X-ray image of the bone structure with enhanced development.
7. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the bone image enhancement method according to any one of claims 1 to 5.
8. A storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by one or more processors, the one or more processors execute the bone image enhancement method as described in any one of claims 1 to 5.
Citation Information
Patent Citations
Automatic registration positioning method and device for bone joints
CN112184782A
Two-dimensional skeleton segmentation method and device based on DRR simulation data generation
CN113496494A
Generation method of spine extraction model and enhancement method and device of spine definition
CN116912497A
Medical image processing method and device and electronic equipment
CN117314867A
Cited By
Image processing device, image processing method, and program
WO2026211360A1
