CN119887789B

CN119887789B - CT image-based lumbar disc herniation degeneration evaluation method and system

Info

Publication number: CN119887789B
Application number: CN202510392748.9A
Authority: CN
Inventors: 王栋; 戚维辉; 潘浩
Original assignee: Hangzhou Traditional Chinese Medicine Hospital Affiliated To Zhejiang University Of Traditional Chinese Medicine Hangzhou Traditional Chinese Medicine Hospital
Current assignee: Hangzhou Traditional Chinese Medicine Hospital Affiliated To Zhejiang University Of Traditional Chinese Medicine Hangzhou Traditional Chinese Medicine Hospital
Priority date: 2025-03-31
Filing date: 2025-03-31
Publication date: 2025-06-06
Anticipated expiration: 2045-03-31
Also published as: CN119887789A

Abstract

The invention discloses a lumbar disc herniation and degeneration evaluation method and system based on CT images, wherein the method comprises the steps of projecting image data of a three-dimensional lumbar disc region to a multi-view plane according to lumbar CT image data to generate a local feature map containing a local structure of a lumbar disc and a global feature map containing a macroscopic structure of the lumbar disc, extracting structural features of a lumbar disc annulus fibrosus, a nucleus pulposus and a herniation from the local feature map and the global feature map of different views to reconstruct an anatomical structure of the lumbar disc in a three-dimensional space to obtain a geometric deformation model of the lumbar disc, constructing a multi-stage evaluation model of lumbar disc degeneration according to the geometric deformation model, training the multi-stage evaluation model through a deep learning network to perform feature classification and quantization scoring, and evaluating the herniation level. By utilizing the embodiment of the invention, the degeneration state of the lumbar intervertebral disc can be comprehensively and accurately analyzed through the combination of the three-dimensional reconstruction and the deep learning technology.

Description

CT image-based lumbar disc herniation degeneration evaluation method and system

Technical Field

The invention belongs to the technical field of image processing, and particularly relates to a lumbar disc herniation and degeneration evaluation method and system based on CT images.

Background

Lumbar intervertebral discs are an important component of the spine, consisting of the annulus and nucleus pulposus, and play a role in cushioning and supporting. With the influence of factors such as aging, genetic factors, trauma, improper long-term posture or excessive strain, the lumbar intervertebral disc gradually degenerates, and finally, the lumbar intervertebral disc protrusion may be caused. The lumbar disc herniation not only causes pain and limited movement of patients, but also can press nerve roots in severe cases, and causes a series of nerve symptoms such as radiation pain, numbness, muscle weakness and the like of lower limbs.

Currently, the assessment of prolapse of lumbar intervertebral disc degeneration is mostly dependent on observation and physical examination of clinical symptoms, or by using imaging examination means such as MRI (magnetic resonance imaging) and CT (computed tomography). MRI has a higher soft tissue contrast and is commonly used to assess the condition of the lumbar disc, but in some cases CT images are more clear in the display of bone and calcification structures, which can provide a more intuitive demarcation of bone and soft tissue. However, the existing image analysis method focuses on the traditional two-dimensional slice observation, and lacks comprehensive knowledge of the lumbar disc structure in the three-dimensional space. The defect can not fully mine hidden potential characteristics and changes in the image data in the degeneration evaluation, so that accurate judgment of lesion degree is affected, and the effects of early diagnosis and personalized treatment are reduced.

Disclosure of Invention

The invention aims to provide a lumbar disc herniation degeneration evaluation method and system based on CT images, which are used for solving the defects in the prior art, and can comprehensively and accurately analyze the degeneration state of a lumbar disc through the combination of three-dimensional reconstruction and deep learning technologies.

One embodiment of the application provides a lumbar disc herniation degeneration assessment method based on CT images, which comprises the following steps:

Projecting the image data of the three-dimensional lumbar disc region to a multi-view plane according to the lumbar CT image data to generate a local feature map containing a local structure of the lumbar disc and a global feature map containing a macroscopic structure of the lumbar disc;

Extracting structural features of the fibrous ring, nucleus pulposus and protrusion of the lumbar intervertebral disc from the local feature map and the global feature map of different view angles to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space so as to obtain a geometric deformation model of the lumbar intervertebral disc;

and constructing a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model, wherein the multistage evaluation model is trained through a deep learning network so as to perform feature classification and quantitative scoring, and evaluate the outstanding degeneration level.

Optionally, the projecting the image data of the three-dimensional lumbar disc region onto the multi-view plane according to the lumbar CT image data to generate a local feature map including a local structure of the lumbar disc and a global feature map including a macroscopic structure of the lumbar disc includes:

performing three-dimensional reconstruction according to lumbar CT image data to construct a three-dimensional lumbar disc model;

projecting the three-dimensional lumbar disc model onto a preset multi-view plane through a projection algorithm, wherein each view comprises three-dimensional information;

In each view angle, extracting a local feature map containing specific key structures of the lumbar intervertebral disc, and generating a corresponding global feature map under the same view angle so as to capture the macroscopic structure of the whole lumbar intervertebral disc and ensure that all relevant anatomical structures are contained.

Optionally, extracting structural features of the annulus fibrosus, the nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map from different view angles to reconstruct an anatomical structure of the lumbar intervertebral disc in a three-dimensional space to obtain a geometric deformation model of the lumbar intervertebral disc, including:

Applying an image segmentation algorithm to process the local feature map, segmenting the fibrous ring, nucleus pulposus and the protruding structures of the lumbar intervertebral disc from the background, and using morphological operation to clarify the boundaries of different segmentation structures in the global feature map;

Extracting key structural features from different clarified segmented structures, wherein the key structural features comprise the thickness and shape of a fiber ring, the volume and shape of a nucleus pulposus and the size and position of a protruding part;

mapping the extracted key structural features into a three-dimensional space, constructing a three-dimensional anatomical structure model of the lumbar intervertebral disc, smoothing and connecting the extracted feature points by adopting an interpolation method to form a complete geometric shape, reconstructing the appearance of the lumbar intervertebral disc under different degeneration states, correcting the geometric shape of the three-dimensional anatomical structure model by a numerical optimization algorithm, ensuring the consistency of the model and original CT image data, and obtaining a reconstructed three-dimensional anatomical structure model;

performing geometric deformation analysis on the reconstructed three-dimensional anatomical structure model, identifying the change of the shape and the degree of degeneration, and extracting geometric deformation characteristics;

The geometric deformation characteristics are integrated into a geometric deformation model to represent the structural change of the lumbar intervertebral disc under different degeneration states, and the geometric deformation model can reflect the relative positions of all the structures and the change degrees of the relative positions.

Optionally, the constructing a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model includes:

On the basis of a geometric deformation model, a high-order tensor analysis is applied to construct a tensor field of the annulus and the nucleus so as to provide deformation information of each region under different stress amounts, and based on the tensor field, the local deformation characteristics of the lumbar intervertebral disc are analyzed through the region deformation tensor, and the protruding displacement of the nucleus, the stress change of the annulus and the tension distribution between the nucleus and the vertebrae are quantified so as to generate a multidimensional tensor characteristic diagram;

according to the multidimensional tensor feature map and the stress and deformation information contained in the multidimensional tensor feature map, performing self-adaptive texture feature segmentation on the boundary region of the nucleus pulposus and the annulus fibrosus to obtain pathological texture feature distribution of lumbar disc tissues;

And constructing a multistage evaluation model of the lumbar disc degeneration by utilizing pathological texture features, wherein the multistage evaluation model is trained through a deep learning network so as to perform feature classification and quantitative scoring, and evaluate the outstanding degeneration level.

Yet another embodiment of the present application provides a lumbar disc herniation degeneration evaluation system based on CT images, the system comprising:

The projection module is used for projecting the image data of the three-dimensional lumbar disc region to a multi-view plane according to the lumbar CT image data so as to generate a local feature map containing a local structure of the lumbar disc and a global feature map containing a macroscopic structure of the lumbar disc;

The extraction module is used for extracting structural features of the fibrous ring, the nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map from different visual angles so as to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space and obtain a geometric deformation model of the lumbar intervertebral disc;

the evaluation module is used for constructing a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model, wherein the multistage evaluation model is trained through a deep learning network so as to perform feature classification and quantization scoring, and the salient degeneration level is evaluated.

A further embodiment of the application provides a storage medium having a computer program stored therein, wherein the computer program is arranged to perform the method of any of the preceding claims when run.

Yet another embodiment of the application provides an electronic device comprising a memory having a computer program stored therein and a processor configured to run the computer program to perform the method recited in any of the preceding claims.

Compared with the prior art, the lumbar disc herniation and degeneration evaluation method based on the CT image is characterized by comprising the steps of projecting image data of a three-dimensional lumbar disc region to a multi-view plane according to lumbar CT image data to generate a local feature map comprising a local structure of the lumbar disc and a global feature map comprising a macroscopic structure of the lumbar disc, extracting structural features of a lumbar disc annulus, a nucleus pulposus and a herniation from the local feature map and the global feature map of different view angles to reconstruct an anatomical structure of the lumbar disc in a three-dimensional space so as to obtain a geometric deformation model of the lumbar disc, constructing a multi-stage evaluation model of lumbar disc degeneration according to the geometric deformation model, wherein the multi-stage evaluation model is trained through a deep learning network so as to perform feature classification and quantization scoring, and evaluate the degeneration level, so that the degeneration state of the lumbar disc can be comprehensively and accurately analyzed through the combination of three-dimensional reconstruction and deep learning technologies.

Drawings

Fig. 1 is a hardware block diagram of a computer terminal of a lumbar disc herniation and degeneration evaluation method based on CT images according to an embodiment of the present invention;

Fig. 2 is a flowchart of a method for evaluating lumbar disc herniation based on CT images according to an embodiment of the present invention;

fig. 3 is a schematic structural diagram of a lumbar disc herniation and degeneration evaluation system based on CT images according to an embodiment of the present invention.

Detailed Description

The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.

The embodiment of the invention firstly provides a lumbar disc herniation degeneration evaluation method based on CT images, which can be applied to electronic equipment such as computer terminals, in particular to common computers and the like.

The following describes the operation of the computer terminal in detail by taking it as an example. Fig. 1 is a hardware block diagram of a computer terminal according to an embodiment of the present invention, which is a method for evaluating lumbar disc herniation and degeneration based on CT images. As shown in fig. 1, the computer device includes a processor, a memory, and a network interface connected by a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

The non-volatile storage medium may store an operating system and a computer program. The computer program comprises program instructions which, when executed, cause the processor to perform any of a number of CT image-based methods for assessing prolapse of lumbar intervertebral disc.

The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

The internal memory provides an environment for the execution of a computer program in the non-volatile storage medium, which when executed by the processor, causes the processor to perform any one of the CT image-based lumbar disc herniation degeneration assessment methods.

The network interface is used for network communication such as transmitting assigned tasks and the like. It will be appreciated by those skilled in the art that the architecture shown in fig. 1 is merely a block diagram of some of the architecture relevant to the present inventive arrangements and is not limiting as to the computer device to which the present inventive arrangements may be implemented, as a particular computer device may include more or less components than those shown, or may be combined with some components, or may have a different arrangement of components.

It should be appreciated that the Processor may be a central processing unit (Central Processing Unit, CPU), it may also be other general purpose processors, digital signal processors (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. Wherein the general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

Referring to fig. 2, an embodiment of the present invention provides a method for evaluating lumbar disc herniation degeneration based on CT images, which may include the following steps:

S201, projecting image data of a three-dimensional lumbar disc region to a multi-view plane according to lumbar vertebra CT image data to generate a local feature map containing a local structure of the lumbar disc and a global feature map containing a macroscopic structure of the lumbar disc;

according to the method, image data of a three-dimensional lumbar intervertebral disc region is projected to a plurality of view planes according to lumbar vertebra CT image data, so that a local feature map containing a local structure of the lumbar intervertebral disc and a global feature map reflecting a macroscopic structure are generated. By means of the projection technology, abundant structural information can be obtained from different view angles, fine local features such as specific shapes and boundaries of the annulus and the nucleus pulposus are covered, and the whole anatomical structure is shown, so that a solid foundation is laid for subsequent feature extraction and three-dimensional reconstruction.

The implementation significance of the step is that the advantages of high resolution and multi-angle observation of CT images are fully utilized, a three-dimensional view is formed, and the anatomical features of the lumbar intervertebral disc can be more comprehensively presented. The method is not only beneficial to improving the accuracy of the subsequent algorithm in feature extraction and model reconstruction, but also provides rich information support for evaluating the degeneration state of the lumbar intervertebral disc, thereby promoting the scientization and individuation of clinical diagnosis and treatment.

Specifically, three-dimensional reconstruction can be performed according to lumbar vertebra CT image data, and a three-dimensional lumbar disc model is constructed;

In this step, first, the lumbar CT image data is collected and processed, and the two-dimensional slice data is integrated into a coherent three-dimensional model by an image reconstruction algorithm (e.g., a filter-based reconstruction technique). The model can truly reflect the spatial distribution and structural characteristics of the lumbar intervertebral disc, and a complete three-dimensional image data set is formed, so that subsequent analysis and processing are facilitated. The three-dimensional reconstruction is the basis for obtaining the lumbar disc morphology information, can effectively relieve the limitation of the traditional two-dimensional image analysis, provides the needed space information for deeper structural analysis and degeneration evaluation, and ensures the accuracy and reliability of the model.

In performing three-dimensional reconstruction, it is first necessary to collect CT image data of the lumbar spine, which is typically stored in DICOM format. After loading the CT image dataset, professional image processing software (such as 3D slice or OsiriX) is used, preprocessing is performed, including noise reduction, image enhancement and contrast adjustment. These steps help to improve the quality of the image and ensure the accuracy of the reconstructed model.

Next, using the three-dimensional reconstruction functions in these software, a suitable reconstruction algorithm, e.g. an interpolation-based filter, is selected to process the plurality of slice data. This process involves analyzing the pixel values of each slice and filling in the gaps between slices by interpolation to form continuous volumetric data. The user can adjust the thickness of the slice and the spatial resolution of the reconstruction to optimize the detail and visualization of the final model.

After the reconstruction is completed, the three-dimensional model is exported into a visualization format (e.g., OBJ or STL) for subsequent operations. After the three-dimensional model is built, further detail adjustment and optimization are performed through three-dimensional visualization software (such as Blender or MeshLab) so as to clearly display the anatomical structure of the lumbar intervertebral disc, and a foundation is laid for subsequent analysis.

this step involves projecting the constructed three-dimensional lumbar disc model from a plurality of preset viewing angles (e.g., front, side, and oblique) by projection algorithms in computer graphics (e.g., perspective projection or orthographic projection, or maximum intensity projection, average intensity projection, etc.). The projection process ensures that each view angle can capture three-dimensional information of the model and comprehensively present the complex structure of the lumbar intervertebral disc. By projecting the three-dimensional model onto a plurality of planes, image data at different viewing angles can be generated, and the comprehensiveness of information acquisition is enhanced. This process provides multi-angle information support for subsequent extraction of local and global features, helping to more accurately analyze and reconstruct the anatomy of the lumbar disc.

After the three-dimensional reconstruction is completed, the next step is to project the three-dimensional lumbar disc model to a plurality of preset two-dimensional viewing angle planes. This process uses projection algorithms in computer graphics to create images based on set viewing angles and projection types (e.g., perspective projection or orthographic projection). In this step, the position, direction and projection matrix of the viewing angle first need to be defined to determine how to convert the three-dimensional coordinate system into a two-dimensional plane.

In specific implementation, openGL or similar 3D graphics engines may be used to set view port and perspective parameters of the model, so as to ensure that each view angle correctly reflects all details of the lumbar disc. By appropriate configuration of camera parameters (e.g., view angle, near plane, and far plane), different viewing angles can be obtained to provide more comprehensive information.

After the projection is completed, the generated two-dimensional image is rendered according to the specific setting of each view angle and is output as an image file. These images will serve as the basis for subsequent feature extraction, and may exhibit different structural details during analysis, helping to subsequently extract local and global features.

In this step, key anatomical features of the lumbar disc, such as boundary information of the annulus and nucleus pulposus, are identified and extracted from the projection images generated from each view angle using image processing techniques such as edge detection and region segmentation. And simultaneously, generating a corresponding global feature map, capturing the macroscopic structure of the whole lumbar intervertebral disc, and ensuring that all relevant anatomical structures are taken into consideration. Through the step, the local and global characteristics of the lumbar intervertebral disc can be effectively separated and identified, and a clear structural basis is provided for the subsequent geometric reconstruction. In addition, the integration of local and global features is beneficial to improving the accuracy of the overall model, and ensuring that the degeneration state of the lumbar intervertebral disc can be comprehensively reflected in subsequent analysis.

In this step, for each projection image generated from a different view angle, an image processing technique is used to extract a local feature map and a global feature map. First, key structures of the lumbar disc, such as annulus fibrosus and nucleus pulposus, are identified by image segmentation techniques (such as threshold segmentation, region growing, or segmentation algorithms based on deep learning). In this process, a local feature map is constructed based on specific pixel intensities and shape features, ensuring that a clear structural boundary is extracted.

Then, in order to obtain the global feature map of the macrostructure, the global information under the same view angle is analyzed by using the same image processing flow. At this time, more complex image processing algorithms, such as edge detection algorithms (e.g., canny) and template matching techniques, may need to be applied to accurately capture the overall morphology of the lumbar disc and its interrelationships. This provides comprehensive anatomical information for subsequent model analysis.

Finally, the local feature map is synthesized with the global feature map to ensure that any important anatomical structures are not missed during the analysis. The two types of information are integrated through an image fusion technology to form a comprehensive characteristic diagram, so that a solid foundation is laid for subsequent three-dimensional reconstruction and geometric deformation analysis. Through the steps, comprehensive evaluation of the lumbar intervertebral disc can be realized, and necessary support is provided for analysis of the degree of degeneration.

S202, extracting structural features of the fibrous ring, nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map of different view angles to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space so as to obtain a geometric deformation model of the lumbar intervertebral disc;

In the method, by extracting the structural features of the lumbar intervertebral disc from the local feature map and the global feature map of different visual angles, the key anatomical structures of the lumbar intervertebral disc, including the annulus fibrosus, the nucleus pulposus and the protruding parts, can be accurately analyzed. The process integrates information of different visual angles in the image data, and provides necessary structural basis for subsequent three-dimensional reconstruction. The extracted characteristics can reflect the form, position and size of the structure, and lay a data foundation for understanding the degeneration state of the lumbar intervertebral disc.

By extracting structural features of the lumbar intervertebral disc, detailed anatomical information of the structure in a three-dimensional space can be obtained, so that necessary data support is provided for constructing a geometric deformation model. The model is not only helpful for understanding the morphological change of the lumbar intervertebral disc, but also reflects the biomechanical characteristics of the lumbar intervertebral disc under different degeneration states, and has important practical significance for the establishment of clinical treatment and personalized medical treatment schemes.

Specifically, an image segmentation algorithm can be applied to process a local feature map, segment the fibrous ring, nucleus pulposus and protruding structures of the lumbar intervertebral disc from the background, and in the global feature map, use morphological operation to clarify the boundaries of different segmentation structures;

In this step, the application of the image segmentation algorithm can effectively identify and extract the key structures of the lumbar intervertebral disc in the image. By applying the techniques of threshold segmentation, region growing, etc., the annulus fibrosis, nucleus pulposus and the herniation can be accurately separated from a complex background. Subsequently, morphological operations such as expansion and corrosion are employed to clarify the boundaries, making the separation between different structures more evident, and allowing for subsequent feature extraction. Through efficient image segmentation and boundary definition, details of an anatomical structure can be accurately extracted, and a reliable basis is provided for subsequent reconstruction work. The method is important for researching the degeneration mechanism and biomechanical characteristics of the lumbar intervertebral disc, and is helpful for making diagnosis and treatment schemes.

In this step, the local feature map needs to be preprocessed first to improve the image quality and lay a foundation for the subsequent image segmentation. The noise may be removed by means of gaussian filtering or median filtering while enhancing the image contrast to highlight the different structures. Subsequently, an appropriate image segmentation algorithm is selected. Common methods include threshold-based Otsu segmentation or semantic segmentation using deep learning models such as U-Net, both of which can effectively separate out the annulus, nucleus and herniated structures. When performing segmentation, reasonable thresholds or training models need to be set so that regions of interest can be accurately identified.

After segmentation is completed, the resulting structure tends to have irregular edges and possible artifacts, and thus morphological operations need to be used to further clarify the boundaries of the different segmented structures. In particular, morphological techniques such as open and closed operations can be employed to eliminate small noise points and smooth structure edges. The open operation may be used to remove small isolated regions, while the closed operation may fill small holes, thereby making the segmented regions smoother and more coherent. Finally, after the processing flows, important structure boundaries in the local feature map are more clear, and a solid foundation is laid for subsequent feature extraction.

In this step, the extraction of key structural features is the core of analysis of the anatomy of the lumbar disc. By further analysis of the segmented structure, important information can be obtained, such as the thickness and shape of the annulus, the volume and morphology of the nucleus, and the size and location of the protrusions. The process involves geometric analysis of the structure, quantization of geometric features such as boundary points, centroids and the like, and accurate extraction of the features is ensured. The extracted key structural features can provide necessary parameters for the geometric deformation analysis of the lumbar intervertebral disc. The information not only can help identify biomechanical changes under different degeneration states, but also provides an important quantitative basis for subsequent model construction, and is helpful for better understanding of the degeneration mechanism of the lumbar intervertebral disc.

At this stage, feature extraction is first required for the clarified segmentation result, and specific operations include using contour detection techniques to obtain boundaries of each structure. The boundaries of the annulus, nucleus and herniation can be quickly found out by using a Canny edge detection algorithm, and combined with the original segmentation result, and specific geometric characteristics are obtained through regional attribute analysis. For example, the thickness of the annulus can be calculated by calculating the shortest distance between any two points on its boundary, and at the same time calculating its shape characteristics, such as area, circumference, shape index, etc., which can quantify the geometric characteristics of the annulus and reflect its stability and health.

For feature extraction of the nucleus pulposus, the volume and morphological features thereof are of great concern. By three-dimensional volume calculation, the segmented region of the nucleus pulposus can be regarded as a three-dimensional body and its volume calculated. When the shape of the nucleus pulposus is further analyzed, important parameters such as length-diameter ratio, surface smoothness and the like can be extracted, and the characteristic indexes can reflect the degeneration condition of the nucleus pulposus. Analysis of the protrusion focuses on its size and location, and by calculating the height, width, and coordinates in three dimensions of the protrusion, the severity of the protrusion can be effectively assessed. The extraction of the features not only provides data support for the subsequent geometric modeling, but also is an important basis for clinical diagnosis.

Finally, all extracted key structural features are required to be arranged into a structured data form, so that subsequent analysis and modeling are facilitated. This may be accomplished by creating a database in which each structural feature is recorded in tabular form, with numerical and descriptive statistics for each feature attached to facilitate subsequent analysis. When the model construction stage begins, the extracted features are directly input into corresponding algorithms, and sufficient basic data is provided for subsequent three-dimensional reconstruction and geometric deformation analysis. At the same time, the integration of the data also helps to improve the accuracy and effectiveness of the subsequent models.

In this step, the extracted key structural features are mapped into a three-dimensional space, and a three-dimensional anatomical model of the lumbar disc is constructed. And smoothly connecting gaps among the characteristic points through an interpolation method, and forming a complete model while guaranteeing geometric continuity. In addition, numerical optimization algorithms (e.g., least squares) are used to adjust the model to be highly consistent with the original CT image data, ensuring biological authenticity and physical usability of the final model. By constructing an accurate three-dimensional anatomical structure model, basic data can be provided for subsequent geometric deformation analysis and biomechanical simulation. The model not only can well reflect the actual appearance of the lumbar intervertebral disc, but also can provide important references for clinical evaluation and the design of personalized treatment schemes, and is beneficial to realizing more effective pathological analysis and intervention measures.

The first step of this step is to three-dimensionally map the extracted key structural features. By defining the position coordinates of each key feature, and reconstructing an anatomical structure model of the lumbar intervertebral disc in a three-dimensional coordinate system. In the mapping process, the accuracy of the relative positions of the feature points needs to be considered, so that the true anatomical features can be reflected. In this process, preliminary geometries can be formed by defining the connection between points by means of computer graphics techniques. In the process of initially constructing the three-dimensional model, delaunay triangulation and other methods can be used to ensure the smoothness of connection and the accuracy of the model.

Then, in order to make the three-dimensional model smoother and more natural, interpolation methods are employed to further process the connection between the feature points. Interpolation methods may include bilinear interpolation or spline interpolation, etc., which are effective to generate intermediate points so as to fill in gaps between feature points and form a continuous geometry. In the process, the interpolation algorithm not only improves the smoothness of the model, but also can better reflect the morphological changes of the lumbar intervertebral disc under different degeneration states. The visual effect of the model is obviously improved, so that subsequent analysis and observation become more visual.

In order to ensure the accuracy of the model, the final step is to correct the geometry by a numerical optimization algorithm. And (3) calculating the difference between the model and the image data through comparison with the original CT image data, and adjusting the model parameters by using a least square method or other optimization algorithms. The process aims to eliminate shape deviations possibly existing in the model and ensure that the actual anatomical structure can be accurately reflected. After numerical optimization, the obtained three-dimensional anatomical structure model not only can provide effective support for clinical research, but also can lay an important foundation for subsequent geometric deformation analysis, thereby improving the reliability and effectiveness of the research.

In this step, the reconstructed three-dimensional anatomical model will be used for geometric deformation analysis. By comparing models under different degeneration states, we can identify the shape change and degree of degeneration of the lumbar disc. This analysis involves not only quantitative calculation of the model geometry, but also comparative analysis in order to clarify the structural differences between the different degeneration phases. The geometrical deformation analysis can reveal the rule and mechanism of the degeneration of the lumbar intervertebral disc, and provide scientific basis for clinical intervention. Through the recognition of the shape change, doctors can be helped to make more accurate diagnosis and treatment schemes, and further the medical effect and life quality of patients are improved.

The primary task of performing geometric deformation analysis is to perform detailed shape recognition on the reconstructed three-dimensional anatomical model. This process involves various measurements and analyses of the geometric properties of the model, focusing on its shape changes, volume reductions, and thickness differences. The behavior of the model under various stress conditions can be simulated by adopting a Finite Element Analysis (FEA) method, and the response of the model under different load and stress conditions is observed by applying biomechanical stress to the model, so that the key characteristics of the model in the deformation process are extracted.

In the dynamic analysis process, the relative deformation and the degree of degeneration of each structure in the model can be identified. The method can be realized by calculating local strain, displacement, geometric indexes and the like. For example, the thickness variation of the annulus, the herniation of the nucleus, and the morphology variation at the herniation can be emphasized to form a series of degeneration evaluation indicators. Quantification of these features not only provides data support for the assessment, but also can provide important basis for the physician in clinical decision. Through the analysis, the biomechanical characteristics of the lumbar intervertebral disc and the change conditions of the lumbar intervertebral disc in different degeneration stages can be comprehensively known.

And finally, integrating the extracted geometric deformation characteristics into a geometric deformation model to comprehensively represent the structural change of the lumbar intervertebral disc under different degeneration states. This not only provides specific geometric deformation information, but also reflects the relative positions of the various structures and their degree of variation. The result of the comprehensive analysis can provide important basis for subsequent clinical treatment and research and clear guidance for doctors. Meanwhile, the establishment of the geometric deformation model provides a good theoretical basis for future related researches, and promotes the deep research on the lumbar disc degeneration.

In this final step, the previously extracted geometric deformation features are integrated into the geometric deformation model to form a comprehensive model for representing the structural changes of the lumbar disc in different degenerative states. The model not only demonstrates the relative position between the various anatomical structures (e.g., annulus and nucleus), but also visually presents the degree of change. The process needs to use geometric modeling and data fusion technology to match and integrate the characteristic data so as to generate a complete geometric deformation model capable of reflecting different degeneration states.

By integrating the geometric deformation characteristics, the formed geometric deformation model can more comprehensively reveal the degeneration mechanism of the lumbar intervertebral disc. The model provides an intuitive tool for the evaluation of clinicians, is helpful for observing and analyzing the degree of degeneration in imaging, and provides scientific support for the establishment and effect evaluation of subsequent treatment schemes. In addition, the model can be used for teaching and research to improve understanding of lumbar disc degeneration.

The work at this stage is focused on integrating the extracted geometric deformation features into the geometric deformation model to construct a complete model reflecting the lumbar intervertebral disc in different degeneration states. First, the features extracted from the phase of the degradation analysis (such as strain, displacement, thickness variation, etc.) need to be entered into the geometric deformation model in a structured manner and defined in relation to the various anatomies of the model. This process requires that the model accurately represent the relative position, geometry and stress distribution of each structure in different states.

To achieve this goal, standardized feature mapping methods may be employed, such as using Principal Component Analysis (PCA) to reduce the dimensions of the high-dimensional features, simplifying the complexity of the model, making it easier to understand. Meanwhile, a corresponding visual interface can be created by means of a graphic visual technology, so that doctors and researchers can intuitively see the change of the lumbar intervertebral disc under different degeneration states. These visualizations can not only help the clinician to better understand the pathology, but can also provide a clearer interpretation for the patient, thereby enhancing the patient's confidence in the treatment regimen.

Finally, the integrated geometric deformation model should be predictive. For this purpose, model verification and adjustment can be performed based on existing degeneration data and the theory of biomechanical models. The model is trained by using the existing clinical data through deep learning or machine learning algorithms, so that the model can be accurately predicted in a new clinical scene. The model has practical significance, has a guiding effect on the degeneration evaluation of the lumbar intervertebral disc, lays a good foundation for future research and clinical application, and improves the scientificity of health management of the lumbar intervertebral disc.

S203, constructing a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model, wherein the multistage evaluation model is trained through a deep learning network so as to perform feature classification and quantization scoring, and evaluating the salient degeneration level.

The construction of a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model is an effective means, and the difference between different degeneration levels can be identified by carrying out deep analysis on the geometric deformation of the lumbar disc. The process trains the extracted features by using a deep learning network, thereby establishing a complex evaluation model. The model can automatically identify and classify the degeneration characteristics of the lumbar intervertebral disc, and realize quantitative scoring. The evaluation not only improves the accuracy of diagnosis, but also provides scientific basis for the formulation of clinical treatment schemes.

The construction of the multistage evaluation model has important clinical application value and academic research significance. Firstly, it provides a quantitative evaluation system, which is helpful for clinicians to formulate more accurate treatment strategies for patients with lumbar disc herniation. And secondly, the accuracy and the reliability of the model can be improved through continuous deep learning training, so that the correlation research between the imaging analysis and the pathological features is promoted, and the deep understanding of the pathological mechanism of the degeneration in the medical field is promoted.

Specifically, a high-order tensor analysis can be applied to construct a tensor field of the annulus and the nucleus on the basis of a geometric deformation model so as to provide deformation information of each region under different stress amounts, and based on the tensor field, the local deformation characteristics of the lumbar intervertebral disc are analyzed through the region deformation tensor, and the protrusion displacement of the nucleus, the stress change of the annulus and the tension distribution between the nucleus and the vertebrae are quantified so as to generate a multidimensional tensor characteristic diagram;

The step constructs tensor fields of the annulus fibrosus and the nucleus pulposus through a high-order tensor analysis technology, and aims to quantify strain behaviors of the lumbar intervertebral disc under different loading conditions. By analyzing the tensor field of each region, the deformation characteristics of the local tissue under the action of external force can be more accurately understood, so that reliable data can be provided for subsequent evaluation. By quantifying the local deformation characteristics, the physician can more clearly understand the extent of the lesion and its effect on surrounding tissue. The process not only helps to improve the objectivity of diagnosis, but also provides an important basis for the selection of treatment schemes.

On the basis of a geometric deformation model, a high-order tensor analysis is firstly required to be applied to construct a tensor field of the annulus and the nucleus pulposus. This process starts with a detailed analysis of the geometric deformation model of the lumbar disc. In the process, a tensor field describing the lumbar disc structure is constructed by selecting a proper high-order tensor library so as to accurately reflect the deformation information of each region under different stress amounts. Specifically, the fiducial points and salient feature points of the tensor field are first determined so that the response of each tissue can be clearly recorded when different biomechanical loads are applied. And then, strain and displacement data under various load conditions are obtained through a numerical simulation technology, and the data are converted into tensor forms, so that a multidimensional tensor characteristic diagram aiming at the lumbar intervertebral disc is formed.

After obtaining the tensor field, the local deformation characteristics of the lumbar disc are analyzed based on the tensor field. The process utilizes a region deformation tensor method to analyze different regions of the nucleus pulposus and the annulus fibrosus one by one so as to quantify the protruding displacement of the nucleus pulposus, the stress change of the annulus fibrosus and the tension distribution between the nucleus pulposus and the vertebrae. The analysis is not only helpful for revealing the stress state of the lumbar intervertebral disc, but also can accurately describe the interrelation among the structures, and form multi-level evaluation of deformation characteristics of each level.

And finally, integrating the analysis results to generate a comprehensive multidimensional tensor feature map so as to represent the geometric deformation features of the lumbar intervertebral disc under different degeneration states. The characteristic map contains deformation information of the annulus and the nucleus pulposus and relative position change of the annulus and the nucleus pulposus, and intuitively displays dynamic response of the lumbar intervertebral disc under the stress. The tensor feature map provides a solid foundation for the construction of a subsequent multi-level assessment model, and ensures that the subsequent analysis has enough details and accuracy.

This step aims to locate and segment the boundary region between the nucleus and annulus by the rich information provided by the multi-dimensional tensor feature map. By the self-adaptive texture feature segmentation method, pathological features of local tissues can be identified more accurately, so that deep analysis of lumbar disc degeneration is realized. The process remarkably improves the visualization degree of the histopathological changes, provides clearer anatomic and pathological relations for clinicians, and is beneficial to improving the diagnosis precision and establishing personalized treatment schemes.

In this step, a detailed analysis of the multidimensional tensor feature map is first required to extract the key stress and deformation information. The core of this process is to understand deeply the physical properties and texture features of each region in the image, so we use various image processing algorithms for feature extraction. Firstly, the visualization effect of the feature map is improved by using local contrast enhancement and other technologies, so that the fine tissue structure is more obvious. Next, depending on the texture features of the image, a suitable adaptive segmentation algorithm is selected, which typically includes a thresholding or clustering-based approach to accurately segment the boundary region of the nucleus and annulus from the background region.

In the process of dividing the boundary region, a morphological processing method is used for further sharpening the dividing result so as to eliminate noise and artifacts and ensure the accuracy of the dividing contour. This process not only helps to improve the accuracy of the segmentation boundary, but also enhances the consistency of the segmented features. By this process, each boundary region is obtained that accurately captures the structural features of the annulus and nucleus pulposus.

Next, pathological texture features of the lumbar intervertebral disc are extracted from the clarified segmentation result. These features include the thickness, shape and relative positional changes of the annulus and nucleus pulposus, which are critical to understanding the pathological changes that occur in the degenerative state of the lumbar disc. Finally, summarizing the extracted pathological texture features to form a comprehensive texture feature distribution map, and providing reliable basic data for a subsequent multi-stage evaluation model, so that the pathological condition of the lumbar intervertebral disc can be scientifically and systematically analyzed and evaluated.

In this step, the pathological texture features are used to construct a multi-level assessment model that, in combination with deep learning techniques, enables a comprehensive assessment of the degenerative state of the lumbar disc. The extracted features can be classified and quantized by training the deep learning network, so that the reflection of different degeneration levels is realized. The multi-level assessment model provides a quantitative tool, so that clinicians can formulate more effective treatment strategies based on objective data, and a foundation is laid for future related researches.

When a multistage evaluation model is constructed, firstly, the extracted pathological texture features are required to be processed in a sorting and standardization mode so as to adapt to the input requirements of a subsequent deep learning network. Specifically, each feature is converted into a numerical format, and normalization processing is performed to prevent the number difference between features from affecting the training of the model. Meanwhile, in order to ensure the diversity and sufficiency of the training data, the training set can be expanded by a data enhancement technique, such as rotation, flipping or adding noise, so as to enhance the universality and robustness of the model.

Next, a deep learning network architecture suitable for lumbar disc degeneration assessment is selected. In general, convolutional Neural Networks (CNNs) are suitable because they have good feature extraction capabilities when processing image data. In network architecture design, the number of network layers and the number of neurons need to be reasonably configured according to the complexity of the features and the capacity of the model. In addition, appropriate activation functions and loss functions are selected to optimize the learning process of the network. In the model training process, the network weight is adjusted through a back propagation algorithm, so that the characteristics can be effectively classified and identified.

And finally, evaluating and predicting new input data through a trained deep learning model. According to the output of the model, the degeneration level of the lumbar intervertebral disc can be quantitatively scored to form a standardized evaluation report. The score not only can reflect the health condition of the lumbar intervertebral disc, but also provides decision support for clinicians. In order to ensure the accuracy of the model, the model can be periodically verified and updated to adapt to new clinical data and research progress, thereby improving the reliability and applicability of the evaluation system.

According to the lumbar vertebra CT image data, the image data of the three-dimensional lumbar intervertebral disc area is projected to a multi-view plane to generate a local characteristic image containing a local structure of the lumbar intervertebral disc and a global characteristic image containing a macroscopic structure of the lumbar intervertebral disc; the method comprises the steps of extracting structural features of the fibrous ring, the nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map from different visual angles to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space to obtain a geometric deformation model of the lumbar intervertebral disc, and constructing a multi-stage evaluation model of lumbar disc degeneration according to the geometric deformation model, wherein the multi-stage evaluation model is trained through a deep learning network to perform feature classification and quantization scoring to evaluate the protrusion degeneration level, so that the degeneration state of the lumbar intervertebral disc can be comprehensively and accurately analyzed through the combination of the three-dimensional reconstruction and the deep learning technology.

Yet another embodiment of the present invention provides a lumbar disc herniation degeneration evaluation system based on CT images, see fig. 3, which may include:

The projection module 301 is configured to project image data of a three-dimensional lumbar disc region onto a multi-view plane according to lumbar CT image data, so as to generate a local feature map including a local structure of the lumbar disc and a global feature map including a macro structure of the lumbar disc;

the extraction module 302 is configured to extract structural features of the lumbar disc annulus, the nucleus pulposus and the protrusion from the local feature map and the global feature map from different perspectives, so as to reconstruct an anatomical structure of the lumbar disc in a three-dimensional space, and obtain a geometric deformation model of the lumbar disc;

the evaluation module 303 is configured to construct a multistage evaluation model of the lumbar disc degeneration according to the geometric deformation model, wherein the multistage evaluation model is trained through a deep learning network to perform feature classification and quantization scoring, and evaluate the level of the protrusion degeneration.

The embodiment of the invention also provides a storage medium, in which a computer program is stored, wherein the computer program is configured to perform the steps of any of the method embodiments described above when run.

Specifically, in the present embodiment, the above-described storage medium may be configured to store a computer program for executing the steps of:

The present invention also provides an electronic device comprising a memory having a computer program stored therein and a processor arranged to run the computer program to perform the steps of any of the method embodiments described above.

Specifically, the electronic apparatus may further include a transmission device and an input/output device, where the transmission device is connected to the processor, and the input/output device is connected to the processor.

Specifically, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:

The construction, features and effects of the present invention have been described in detail with reference to the embodiments shown in the drawings, but the above description is only a preferred embodiment of the present invention, but the present invention is not limited to the embodiments shown in the drawings, all changes, or modifications to the teachings of the invention, which fall within the meaning and range of equivalents are intended to be embraced therein, are intended to be embraced therein.

Claims

1. A method for evaluating prolapse of lumbar intervertebral disc based on CT images, comprising:

Extracting structural features of the fibrous ring, the nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map of different visual angles to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space to obtain a geometric deformation model of the lumbar intervertebral disc, wherein an image segmentation algorithm is applied to process the local feature map, the fibrous ring, the nucleus pulposus and the protrusion of the lumbar intervertebral disc are segmented from the background, and boundaries of different segmentation structures are clarified in the global feature map by morphological operation;

The method comprises the steps of carrying out geometric deformation analysis on a reconstructed three-dimensional anatomical structure model, identifying the change and the degree of degeneration of the shape, and extracting geometric deformation characteristics, integrating the geometric deformation characteristics into the geometric deformation model to represent the structural change of the lumbar intervertebral disc under different degeneration states, wherein the geometric deformation model can reflect the relative position and the change degree of each structure;

2. The method of claim 1, wherein projecting the image data of the three-dimensional lumbar disc region onto the multi-view plane based on the lumbar CT image data to generate a local feature map comprising a local structure of the lumbar disc and a global feature map comprising a macro structure of the lumbar disc comprises:

3. The method of claim 2, wherein constructing a multi-level assessment model of lumbar disc degeneration from the geometric deformation model comprises:

4. A lumbar disc herniation degeneration assessment system based on CT images, the system comprising:

The device comprises an extraction module, a global feature map, a segmentation module and a segmentation module, wherein the extraction module is used for extracting structural features of the fibrous ring, the nucleus pulposus and the protrusion of the lumbar intervertebral disc from the local feature map and the global feature map from different visual angles so as to reconstruct the anatomical structure of the lumbar intervertebral disc in a three-dimensional space to obtain a geometric deformation model of the lumbar intervertebral disc;

5. The system according to claim 4, wherein the projection module is specifically configured to:

6. The system according to claim 5, wherein the evaluation module is specifically configured to:

7. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of claims 1-3.