Intelligent quantitative evaluation system for teenager scoliosis fused with multi-modal imaging

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a teenager scoliosis intelligent quantitative evaluation system fused with multi-modal imaging. The system comprises a multi-modal data management module, a cross-modal registration fusion module, a dynamic quantitative analysis module, a risk prediction module, a three-dimensional reconstruction module, a report generation module and a system management module, and by setting a multi-modal registration verification mechanism, accurate matching of a bony structure and soft tissue is ensured. Key parameter calculation errors caused by registration failure are reduced, the three-dimensional reconstruction precision is greatly improved compared with traditional single-mode evaluation, a dynamic respiration compensation module is introduced, in the spine deformation parameter quantification stage, spine physiologic changes in a respiration movement period are automatically captured, measured values are corrected based on an individualized dynamic model, and the accuracy of three-dimensional reconstruction is improved. Pathological progression and physiological fluctuation are distinguished, the misjudgment risk caused by neglecting breathing phase change in traditional static measurement is avoided, and the reliability of scoliosis progression risk assessment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and specifically provides an intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging. Background Technique

[0002] Scoliosis is a common spinal deformity disease, also known as spinal curvature. It is mainly manifested as a lateral deviation of the spine to one side, making the back present a C-shaped or S-shaped curve. When the spinal X-ray film shows a lateral curvature of the spine greater than 10°, it can be diagnosed as scoliosis. This disease is prone to occur in adolescence, and the symptoms gradually worsen with age. In severe cases, it affects respiratory and cardiac functions, and even leads to spinal cord compression and paralysis.

[0003] Currently, in the clinical evaluation of adolescent idiopathic scoliosis, the application of multimodal imaging technology has defects: due to the misalignment of the anatomical coordinate systems of different modal images, it is impossible to detect the registration deviation between bone and soft tissue in real time during the three-dimensional spinal structure reconstruction. If the registration error exceeds 3 mm, it will cause the calculation of key parameters to be distorted; at the same time, the existing evaluation system cannot monitor the impact of respiratory movement on spinal deformation in real time, misjudging physiological deformation as pathological progression, resulting in a risk prediction accuracy reduction of more than 20%; in addition, due to the black box characteristics of the deep learning algorithm, it is impossible to visualize the decision-making basis of the model during key point detection, resulting in doctors being unable to verify the reliability of AI markings and causing clinical misdiagnosis risks; during the dynamic quantitative evaluation process, due to the lack of a multi-dimensional parameter fusion mechanism, it is impossible to realize the collaborative analysis of deformation parameters and biomechanical characteristics, further restricting the generation accuracy of personalized treatment plans and affecting the clinical applicability of scoliosis evaluation.

[0004] Therefore, in view of the above problems, the present invention provides an intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging. Summary of the Invention

[0005] (1) Technical Problems to be Solved In view of the deficiencies of the prior art, the present invention provides an intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging, which solves the problems raised in the above background technique.

[0006] (2) Technical Solutions To achieve the above objectives, the present invention provides the following technical solutions: An intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging, including: Multimodal Data Management Module: An image acquisition unit configured with a medical imaging communication protocol interface to obtain X-ray films, CT, and MRI data; A structured database unit for storing clinical parameters and supporting batch import and export; Cross-modal Registration and Fusion Module: An affine transformation unit that aligns the anatomical structures of X-ray - CT. A GAN fusion unit that realizes the spatial domain feature fusion of MRI and CT. Dynamic Quantification Analysis Module: A key point detection unit that integrates an improved real-time multi-person pose estimation network model and a nested U-shaped network. A parameter calculation unit that dynamically calculates the spinal curvature angle, AVR, and sagittal plane balance parameters. Risk Prediction Module: An extreme gradient boosting tree ensemble learning unit that outputs the probability of scoliosis progression. An interpretability processing unit that generates a SHAP feature contribution degree map. 3D Reconstruction Module: A point cloud processing unit that constructs a 3D spinal mesh model. A deformation simulation unit that visualizes the spinal deformation under different loads. Report Generation Module: A multi-view fusion unit that integrates the 3D model, heat map, and time series curve. An early warning trigger unit that generates red, yellow, and blue early warning signals according to the risk level. System Management Module: A permission control unit based on the role-based access control model that differentiates the operation permissions of doctors, patients, and administrators. An audit log unit that records data access and algorithm call behaviors.

[0007] Preferably, it includes the following steps: S1. Collect multi-modal medical image data and clinical parameters of the patient. The multi-modal medical images include spinal X-ray films, computed tomography scans, and magnetic resonance imaging. The clinical parameters include gender, age, skeletal maturity index, spinal curvature angle, and scoliosis classification system classification. S2. Perform cross-modal registration and fusion processing on the multi-modal medical images: Use affine transformation to align the coronal and sagittal anatomical structures of the X-ray film and CT image. Fuse the MRI soft tissue information into the CT bony structure coordinate system based on the generative adversarial network. Generate a standardized 3D spinal fusion model. S3. Perform multi-plane key point detection on the fusion model: Use an improved real-time multi-person pose estimation model in the coronal plane to detect vertebral corner points and endplate marker points. Use a nested U-shaped network in the sagittal plane to identify the sacral tilt angle and the thoracolumbar turning point. S4. Dynamically quantify spinal deformation parameters: Calculate the main curve and secondary curve spinal curvature angles based on the key points in the coronal plane: Where are the normal vectors of the endplates of the upper and lower end vertebrae respectively; Calculate the thoracic kyphosis angle and lumbar lordosis angle based on the key points in the sagittal plane; S5. Construct a multi-factor risk prediction model: Input the clinical parameters and deformation parameters into the Extreme Gradient Boosting integrated learning framework; Analyze the feature contribution degree through the Shapley Additive Explanation value and output the probability of scoliosis progression ; S6. Generate a dynamically quantified evaluation report: Fuse the three-dimensional reconstruction model, key point heat map, and parameter time series comparison curve; Trigger warning signals according to the grading of the progression probability.

[0008] Preferably, the cross-modal registration in step S2 specifically includes: S21. Perform edge enhancement and gray normalization on the X-ray film; S22. Extract the bony landmark points of the CT image using Scale-Invariant Feature Transform; S23. Calculate the projection transformation matrix between the X-ray film and the CT through the Random Sample Consensus algorithm : Where is the affine transformation function, is the coordinate of the i-th bony landmark point in the CT image, is the coordinate of the i-th projection point matched in the X-ray film, is the number of pairs of matched feature points; S24. Use the Cycle Generative Adversarial Network to map the MRI T2-weighted image to the CT coordinate system: Where is the generative adversarial network from MRI to CT, is the original MRI image, is the synthesized pseudo-CT image.

[0009] Preferably, the key point detection in step S3 adopts a cascaded optimization mechanism: S31. Locate the region of interest of the vertebral body through the spatial attention module; S32. Extract the coronal plane key point coordinates using the high-resolution network with adaptive sampling ; S33. Construct the topological relationship constraint of the vertebral body based on the graph convolutional network; S34. Optimize the sagittal plane key point sequence through the bidirectional long short-term memory network .

[0010] Preferably, the parameter calculation in step S4 includes a dynamic compensation mechanism: S41. Segment the CT sequence frames according to the respiratory phase; S42. Calculate the change amount of the spinal curvature angle in each phase ; S43. Establish a deformation-respiratory cycle correlation function: where is the deformation amplitude coefficient, is the respiratory phase, is the duration of a complete respiratory cycle, is the baseline offset.

[0011] Preferably, the model construction in step S5 includes: S51. Process the unbalanced distribution of clinical data through the synthetic minority over-sampling technique; S52. Use Bayesian optimization to search for the extreme gradient boosting tree hyperparameter combination; S53. Construct a multi-task output layer: Task 1: Scoliosis progression probability ; Task 2: Surgical indication classification label .

[0012] Preferably, it further includes an algorithm adaptive optimization step: S61. Establish an algorithm library: including the key point detection algorithm and the prediction model; S62. Real-time monitor the area under the curve value and F1 score of each algorithm in the validation set; S63. When is less than -σ, trigger the algorithm switch: Use the improved osprey search algorithm to select the optimal alternative algorithm : where is the algorithm library set, is the clinical weight factor, is the area under the curve, is the F1 score.

[0013] Preferably, a three-dimensional reconstruction step is added after step S2: S71. Construct a spine point cloud model based on the registration result; S72. Generate a topologically optimized three-dimensional mesh through Poisson surface reconstruction; S73. Calculate the vertebral rotation degree AVR on the mesh model: where is the average vertebral rotation degree, is the number of vertebrae participating in the calculation, is the normal vector of the cross-section of the i-th vertebra, is the normal vector of the reference plane.

[0014] Preferably, interpretability processing is introduced in step S5: S81. Generate a key point detection heat map through gradient-weighted class activation mapping; S82. Use the local interpretable model-agnostic explanation algorithm to explain the prediction results of the extreme gradient boosting tree; S83. Visualize the feature contribution degree ranking in the evaluation report.

[0015] Preferably, it further includes a cloud platform deployment step: S91. Package the evaluation algorithm using containerization technology; S92. Dynamically allocate computing resources through a container orchestration system; S93. Implement hierarchical permission control for doctors, patients, and administrators based on an open authorization protocol.

[0016] (III) Beneficial effects Compared with the prior art, the present invention provides an intelligent quantitative evaluation system for adolescent idiopathic scoliosis that integrates multi-modal imaging, and has the following beneficial effects: 1. In the present invention, by setting a multi-modal registration verification mechanism, during the fusion process of the spine three-dimensional model, the anatomical structure alignment deviation between X-ray, CT, and MRI images is detected in real time. When the error exceeds the reasonable range, the calibration process is automatically triggered and a spatial error heat map is generated, ensuring the precise matching of bony structures and soft tissues, reducing the calculation error of key parameters caused by registration failure, and significantly improving the three-dimensional reconstruction accuracy compared with traditional single-modal evaluations.

[0017] 2. In the present invention, by introducing a dynamic breathing compensation module, during the quantification stage of spine deformation parameters, the physiological changes of the spine during the breathing movement cycle are automatically captured, and the measured values are corrected based on an individualized dynamic model to distinguish pathological progression from physiological fluctuations, avoiding the risk of misjudgment caused by ignoring the breathing phase change in traditional static measurements, and improving the reliability of scoliosis progression risk assessment.

[0018] 3. In the present invention, by constructing an adaptive algorithm optimization engine, the performance decay of the core algorithm is continuously monitored during the operation of the system. When a decrease in diagnostic accuracy is detected, the optimal alternative algorithm is automatically switched, solving the problem of algorithm failure caused by differences in imaging devices and special body types, and enhancing the system's adaptability to different clinical scenarios.

[0019] 4. In the present invention, by integrating a dual-channel interpretability analysis module, the decision-making process of deep learning is transformed into a visual heat map and a feature contribution map, intuitively showing the generation basis of key parameters in a three-dimensional spine model, enabling doctors to trace the judgment logic of AI and significantly reducing the clinical misdiagnosis risk caused by the black-box nature of the algorithm.

[0020] 5. In the present invention, by establishing a multi-dimensional diagnosis and treatment closed-loop system, integrating biomechanical characteristics, dynamic deformation parameters, and clinical classification data, a quantitative evaluation report and a personalized orthopedic plan are synchronously generated, realizing the full-link optimization from image analysis to clinical decision-making, and improving the diagnosis and treatment efficiency of scoliosis and the accuracy of plan matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a framework diagram of an intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Please refer to Figure 1 , an intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging, comprising: Multi-modal data management module: An image acquisition unit configured with a medical imaging communication protocol interface to obtain X-ray films, CT, and MRI data; A structured database unit for storing clinical parameters and supporting batch import and export; Cross-modal registration and fusion module: An affine transformation unit for performing anatomical structure alignment of X-ray - CT; A GAN fusion unit for realizing spatial domain feature fusion of MRI and CT; Dynamic quantitative analysis module: A key point detection unit integrating an improved real-time multi-person pose estimation network model and a nested U-shaped network; A parameter calculation unit that dynamically calculates the spinal curvature angle, AVR, and sagittal plane balance parameters; A risk prediction module: An extreme gradient boosting tree ensemble learning unit that outputs the probability of scoliosis progression; An interpretability processing unit that generates a SHAP feature contribution map; A three-dimensional reconstruction module: A point cloud processing unit that constructs a three-dimensional spinal mesh model; A deformation simulation unit that visualizes spinal deformations under different loads; A report generation module: A multi-view fusion unit that integrates the three-dimensional model, heat map, and time series curve; An early warning trigger unit that generates red, yellow, and blue early warning signals according to the risk level; A system management module: A permission control unit based on the role-based access control model that distinguishes the operation permissions of doctors, patients, and administrators; An audit log unit that records data access and algorithm call behaviors.

[0024] Including the following steps: S1. Collect multi-modal medical image data and clinical parameters of the patient. The multi-modal medical images include spinal X-rays, computed tomography scans, and magnetic resonance imaging. The clinical parameters include gender, age, skeletal maturity index, spinal curvature angle, and scoliosis classification system classification; S2. Perform cross-modal registration and fusion processing on the multi-modal medical images: Use affine transformation to align the coronal and sagittal anatomical structures of the X-ray and CT images; Fuse MRI soft tissue information into the CT bony structure coordinate system based on the generative adversarial network; Generate a standardized three-dimensional spinal fusion model; S3. Perform multi-plane key point detection on the fusion model: Use an improved real-time multi-person pose estimation model in the coronal plane to detect vertebral corner points and endplate marker points; Use a nested U-shaped network in the sagittal plane to identify the sacral tilt angle and the thoracolumbar turning point; S4. Dynamically quantify spinal deformation parameters: Calculate the main curve and secondary curve spinal curvature angles based on the key points in the coronal plane: Where Are the normal vectors of the upper and lower endplate vertebral bodies respectively; Calculate the thoracic kyphosis angle and lumbar lordosis angle based on the key points in the sagittal plane; S5. Construct a multi-factor risk prediction model: Input clinical parameters and deformation parameters into the extreme gradient boosting integrated learning framework; Analyze the feature contribution degree through the Shapley additive explanation value, and output the scoliosis progression probability ; S6. Generate a dynamic quantization evaluation report: Fuse the three-dimensional reconstruction model, key point heat maps, and parameter time series comparison curves; Trigger warning signals according to the progression probability grading.

[0025] The cross-modal registration in step S2 specifically includes: S21. Perform edge enhancement and grayscale normalization on the X-ray film; S22. Extract the bony landmark points of the CT image using scale-invariant feature transform; S23. Calculate the projective transformation matrix between the X-ray film and the CT through the random sample consensus algorithm : where is the affine transformation function, is the coordinate of the i-th bony landmark point in the CT image, is the coordinate of the i-th projected point matched in the X-ray film, is the number of matched feature point pairs; S24. Use the cycle generative adversarial network to map the MRI T2-weighted image to the CT coordinate system: where is the generative adversarial network from MRI to CT, is the original MRI image, is the synthesized pseudo-CT image.

[0026] The key point detection in step S3 adopts a cascaded optimization mechanism: S31. Locate the region of interest of the vertebral body through the spatial attention module; S32. Extract the coronal plane key point coordinates using the high-resolution network with adaptive sampling ; S33. Construct the vertebral body topological relationship constraint based on the graph convolutional network; S34. Optimize the sagittal plane key point sequence through the bidirectional long short-term memory network .

[0027] The parameter calculation in step S4 includes a dynamic compensation mechanism: S41. Segment the CT sequence frames according to the respiratory phase; S42. Calculate the angular change amount of the spinal curvature angle in each phase ; S43. Establish a deformation - respiratory cycle correlation function: where is the deformation amplitude coefficient, is the respiratory phase, is the duration of a complete respiratory cycle, is the baseline offset.

[0028] The model construction in step S5 includes: S51. Process the unbalanced distribution of clinical data through the Synthetic Minority Over - sampling Technique (SMOTE); S52. Use Bayesian optimization to search for the hyperparameter combination of the Extreme Gradient Boosting (XGBoost) tree; S53. Construct a multi - task output layer: Task 1: Scoliosis progression probability ; Task 2: Surgical indication classification label .

[0029] It also includes an algorithm adaptive optimization step: S61. Establish an algorithm library: including the key - point detection algorithm and the prediction model; S62. Real - time monitor the area under the curve (AUC) value and the F1 - score of each algorithm on the validation set; S63. When is less than -σ, trigger an algorithm switch: Use the improved Osprey search algorithm to select the optimal alternative algorithm : where is the algorithm library set, is the clinical weight factor, is the area under the curve, is the F1 - score.

[0030] Add a 3D reconstruction step after step S2: S71. Construct a spine point - cloud model based on the registration result; S72. Generate a topologically optimized 3D mesh through Poisson surface reconstruction; S73. Calculate the average vertebral rotation (AVR) on the mesh model: where is the average vertebral rotation, is the number of vertebrae involved in the calculation, is the normal vector of the cross-section of the i-th vertebral body, is the normal vector of the reference plane.

[0031] In step S5, interpretability processing is introduced: S81. Generate a heatmap for key point detection through Gradient-weighted Class Activation Mapping; S82. Use the Local Interpretable Model-Agnostic Explanation algorithm to interpret the prediction results of the Extreme Gradient Boosting tree; S83. Visualize the sorted feature contributions in the evaluation report.

[0032] It also includes the steps of cloud platform deployment: S91. Package the evaluation algorithm using containerization technology; S92. Dynamically allocate computing resources through a container orchestration system; S93. Implement hierarchical permission control for doctors, patients, and administrators based on an open authorization protocol.

[0033] Example 1: Multimodal Registration Fusion and 3D Reconstruction During the system implementation process, first collect the patient's spinal X-ray films, CT, and MRI sequences through the DICOM interface. When performing cross-modal registration, use the SIFT algorithm to extract the vertebral body edges and pedicle bony landmark points from the CT images. At the same time, map the MRI T2-weighted images to the CT coordinate system through an improved CycleGAN network to generate pseudo-CT images. When the spatial registration module detects an alignment deviation of 2.3 mm between the sacral endplate plane in CT and MRI, it automatically triggers the re-registration process: recalculate the affine transformation matrix using the RANSAC algorithm, and overlay the registration error on the 3D model as a heatmap for display.

[0034] After registration, the point cloud processing unit fuses multimodal data to construct a 3D spinal mesh model, and generates a topologically optimized digital twin of the spine through Poisson surface reconstruction. The calculation accuracy of the vertebral body rotation degree reaches the sub-millimeter level, significantly eliminating the anatomical structure distortion caused by a single imaging modality.

[0035] Example 2: Dynamic Quantification and Risk Prediction Closed Loop In the dynamic evaluation stage, the system automatically segments the respiratory phases based on the 4D-CT sequence, detects the vertebral corner points on the coronal plane of each phase through the HRNet network with adaptive sampling, and optimizes the sagittal key point sequence in combination with a bidirectional LSTM. The parameter calculation unit applies a dynamic compensation function: real-time correct the measured value of the spinal curvature angle according to the individual respiratory cycle duration of the patient, eliminate the 6.1° pseudo-deviation caused by spinal extension during inspiration, and input the corrected parameters into the XGBoost multi-task model to synchronously process the clinically enhanced database with SMOTE.

[0036] When the main spinal curvature angle of a 15-year-old female patient is detected to be 31° and the sagittal thoracolumbar turning point offset is identified, the system links the biomechanical simulation unit to calculate the intervertebral disc stress distribution, generate the probability of scoliosis progression and surgical indication warning, and simultaneously trigger the three-dimensional correction scheme simulation module to output the personalized brace mechanical parameters.

[0037] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0038] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent quantitative evaluation system for adolescent scoliosis integrating multimodal imaging, characterized in that: Including: Multimodal data management module: An image acquisition unit configured with a medical imaging communication protocol interface to obtain X-ray films, CT, and MRI data; A structured database unit that stores clinical parameters and supports batch import and export; Cross-modal registration and fusion module: An affine transformation unit that performs anatomical structure alignment of X-ray - CT; A GAN fusion unit that realizes spatial domain feature fusion of MRI and CT; Dynamic quantization analysis module: A key point detection unit that integrates an improved real-time multi-person pose estimation network model and a nested U-shaped network; A parameter calculation unit that dynamically calculates spinal curvature angles, AVR, and sagittal balance parameters; Risk prediction module: An extreme gradient boosting tree ensemble learning unit that outputs the probability of scoliosis progression; An interpretability processing unit that generates a SHAP feature contribution degree map; 3D reconstruction module: A point cloud processing unit that constructs a 3D spinal mesh model; A deformation simulation unit that visualizes spinal deformations under different loads; Report generation module: A multi-view fusion unit that integrates 3D models, heat maps, and time series curves; An early warning trigger unit that generates red, yellow, and blue early warning signals according to the risk level; System management module: A permission control unit based on the role-based access control model that differentiates the operation permissions of doctors, patients, and administrators; An audit log unit that records data access and algorithm call behaviors.

2. The intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging according to claim 1, wherein: Including the following steps: S1. Collect the multimodal medical image data and clinical parameters of the patient. The multimodal medical images include spinal X-ray films, computed tomography scans, and magnetic resonance imaging. The clinical parameters include gender, age, skeletal maturity index, spinal curvature angle, and scoliosis classification system classification; S2. Perform cross-modal registration and fusion processing on the multimodal medical images: Use affine transformation to align the coronal and sagittal anatomical structures of the X-ray film and CT image; Fuse MRI soft tissue information into the CT bony structure coordinate system based on the generative adversarial network; Generate a standardized 3D spinal fusion model; S3. Perform multi-plane key point detection on the fusion model: Use an improved real-time multi-person pose estimation model in the coronal plane to detect vertebral corner points and endplate marker points; Use a nested U-shaped network in the sagittal plane to identify the sacral tilt angle and the thoracolumbar turning point; S4. Dynamically quantify spinal deformation parameters: Calculate the main curve and secondary curve spinal curvature angles based on the key points in the coronal plane; Among them are the normal vectors of the upper and lower endplate vertebral endplates respectively; Calculate the thoracic kyphosis angle and lumbar lordosis angle based on the key points in the sagittal plane; S5. Construct a multi-factor risk prediction model: Input the clinical parameters and deformation parameters into the extreme gradient boosting ensemble learning framework; Parse the feature contribution degree through the Shapley additive explanation value and output the probability of scoliosis progression ; S6. Generate a dynamic quantization evaluation report: Fuse the 3D reconstruction model, key point heat map, and parameter time series comparison curve; Trigger early warning signals according to the progression probability grading.

3. The intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging according to claim 2, characterized in that: The cross-modal registration in step S2 specifically includes: S21. Perform edge enhancement and gray normalization on the X-ray film; S22. Use scale-invariant feature transform to extract bony landmark points of the CT image; S23. Calculate the projection transformation matrix between the X-ray film and the CT by the Random Sample Consensus algorithm : wherein is an affine transformation function, is the coordinate of the i-th bony landmark point in the CT image, is the coordinate of the i-th projection point matched in the X-ray film, is the number of matched feature point pairs; S24. Use a cyclic generative adversarial network to map the MRI T2-weighted image to the CT coordinate system: Among them is the generative adversarial network from MRI to CT, is the original MRI image, is the synthetic pseudo-CT image.

4. The intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging according to claim 2, characterized in that: The key point detection in step S3 adopts a cascaded optimization mechanism: S31. Locate the region of interest of the vertebral body through a spatial attention module; S32. Extract the coronal plane key point coordinates using an adaptive sampling high-resolution network ; S33. Construct the topological relationship constraint of the vertebral body based on the graph convolutional network; S34. Optimize the sagittal plane key point sequence through a bidirectional long short-term memory network .

5. The intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging according to claim 2, characterized in that: The parameter calculation in step S4 includes a dynamic compensation mechanism: S41. Segment the CT sequence frames according to the respiratory phase; S42. Calculate the angular change amount of the spinal curvature angle for each phase ; S43. Establish a deformation-respiratory cycle correlation function: Among them is the deformation amplitude coefficient, is the respiratory phase, is the duration of a complete respiratory cycle, is the baseline offset.

6. The intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging according to claim 2, characterized in that: The model construction in step S5 includes: S51. Process the unbalanced distribution of clinical data through the synthetic minority over-sampling technique; S52. Use Bayesian optimization to search for the hyperparameter combination of the extreme gradient boosting tree; S53. Construct a multi-task output layer: Task 1: Probability of Scoliosis Progression ; Task 2: Classification Labels for Surgical Indications 。 7. An intelligent quantitative evaluation system for adolescent scoliosis integrating multi-modal imaging according to claim 2, characterized in that: It also includes an algorithm adaptive optimization step: S61. Establish an algorithm library: It includes a key point detection algorithm and a prediction model; S62. Real-time monitor the area under the curve value and F1 score of each algorithm on the validation set; S63. When is less than -σ, trigger algorithm switching: Select the optimal alternative algorithm using the improved osprey search algorithm : Among them is the algorithm library set, is the clinical weight factor, is the area under the curve, is score.

8. The intelligent quantitative evaluation system for adolescent scoliosis integrating multimodal imaging according to claim 2, wherein: A three-dimensional reconstruction step is added after step S2: S71. Construct a spine point cloud model based on the registration result; S72. Generate a topologically optimized three-dimensional mesh through Poisson surface reconstruction; S73. Calculate the vertebral rotation degree AVR on the mesh model; wherein is the average vertebral rotation degree, is the number of vertebrae involved in the calculation, is the normal vector of the cross-section of the i-th vertebra, is the normal vector of the reference plane.

9. The intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multi-modal imaging according to claim 2, wherein: Interpretability processing is introduced in step S5: S81. Generate a key point detection heat map through gradient-weighted class activation mapping; S82. Use the local interpretable model-agnostic explanation algorithm to explain the prediction results of the extreme gradient boosting tree; S83. Visualize the feature contribution degree ranking in the evaluation report.

10. An intelligent quantitative evaluation system for adolescent idiopathic scoliosis integrating multimodal imaging, characterized in that: It also includes a cloud platform deployment step: S91. Package the evaluation algorithm using containerization technology; S92. Dynamically allocate computing resources through a container orchestration system; S93. Implement hierarchical permission control for doctors, patients, and administrators based on an open authorization protocol.

Images