Image recognition system for constructing osteoporotic vertebral fracture healing based on artificial intelligence

Abstract

The invention discloses an image recognition system for constructing osteoporotic vertebral fracture healing based on artificial intelligence, and particularly relates to the technical field of orthopedic treatment, which comprises a data preprocessing module for extracting and optimizing data from an original medical image. According to the method, the advantage of discipline crossing is developed, a recognition system on medical imaging is developed based on deep learning, the advantage that the recognition system can discover detail features and hidden rules which cannot be found by human beings is utilized, and then a large amount of structured data is utilized to automatically learn and train visual feature expression of abstract data, so that the clinical missed diagnosis rate is reduced, and the clinical diagnosis efficiency is improved. The artificial intelligence architecture is trained by a rigorous and scientific means, so that the artificial intelligence architecture has clinical experience and knowledge of high-age dominant doctors and is popularized in local hospitals to achieve an auxiliary diagnosis effect, the difference of medical differentiation is reduced, the technical and experience defects of local medical personnel are made up, and the technical and experience defects of the local medical personnel are overcome. The aim of improving the definite diagnosis rate of diseases and reducing the missed diagnosis rate is one of the purposes of the invention.

Description

Technical Field

[0001] The present invention relates to the technical field of orthopedic treatment, and in particular to an image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence. Background Art

[0002] With the advent of aging, osteoporotic vertebral fractures are also increasing. The clinical work of orthopedic hospitals mainly includes the following aspects: diagnosis and evaluation, surgical treatment, follow-up and monitoring, multidisciplinary collaboration, patient education, scientific research and teaching. Through these works, the quality of life and rehabilitation effect of patients are improved. The current treatment options are mainly conservative treatment and surgical treatment, and the surgery mainly uses vertebral reinforcement. However, there is still a lack of authoritative guidelines for treatment indications. Therefore, there is controversy in the choice of treatment for osteoporotic vertebral fractures, and patients are also at risk of nonunion of fractures. Nonunion of fractures will lead to irreversible long-term chronic low back pain. When there is an unstable fracture, it may lead to complications such as neurological symptoms and kyphosis. When conservative treatment is ineffective and surgical intervention is required, for patients with deformities, nerve damage or disabilities, the surgical risks and costs increase exponentially, and the recovery effect after surgery is often unsatisfactory. Therefore, we provide an image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence. Summary of the invention

[0003] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: An image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence, including a data preprocessing module for extracting and optimizing data from original medical images, a feature extraction and model building module for automatically extracting features of fracture sites through deep learning technology and building a suitable recognition model, a bone healing monitoring and classification module for classifying the healing status of identified fracture images and tracking the fracture healing process in real time, and a system optimization and scalability module for improving the operating efficiency and scope of application of the system to ensure that it can be applied in medical institutions at all levels; The data preprocessing module includes an image denoising module for removing noise or artifacts in the image, an image enhancement module for improving the visibility of the fracture site by adjusting contrast, brightness, and sharpening, an image segmentation module for separating the vertebral body from the surrounding tissue using U-Net segmentation technology, and a data expansion module for generating more samples by methods such as rotation, scaling, and translation; The feature extraction and model construction module includes a feature point detection module for extracting important feature points of fracture edges and bone density using traditional SIFT algorithms or deep learning methods, a multi-scale feature fusion module for combining global and local features of the vertebral body by introducing convolution kernels of different scales, a convolutional neural network construction module for building a lightweight CNN architecture, and an attention mechanism integration module for focusing more attention of the model on the fracture healing site using the attention mechanism.

[0005] The present invention is further configured as follows: the bone healing monitoring and classification module includes a bone healing status evaluation module for automatically classifying the fracture healing stage based on indicators such as bone density and fracture line healing conditions, a multi-time point tracking and comparison module for tracking the healing progress by combining imaging data of patients at different times to establish a time series model, an anomaly detection module for automatically issuing early warnings when anomalies occur during the fracture healing process, and an expert-assisted treatment module for providing decision support to primary care physicians to recommend further examinations or treatment plans. The system optimization and scalability module includes a hardware compatibility optimization module for optimizing the hardware requirements of the system for the device performance of different levels of medical institutions so that it can operate in environments with limited device conditions such as county-level hospitals or township health centers, a model lightweight module for reducing the complexity and computational amount of the model through techniques such as model pruning and quantization to ensure that it can run in real time on ordinary devices, a federated learning integration module for ensuring secure data sharing between different medical institutions through federated learning technology, and a remote update and maintenance module for providing the system with a remote update function to facilitate the continuous optimization and upgrade of the model by the development team.

[0006] The present invention is further configured as follows: The multi-time-point tracking and comparison module includes a time-series data aggregation module for aggregating and standardizing the image data obtained by the patient at different time points to ensure the consistency of data format and resolution, a dynamic feature extraction module for extracting dynamic features such as bone density and healing rate that change over time from the time-series images, a feature fusion trend prediction module for fusing the extracted dynamic features to generate a healing trend curve to predict the possible future development direction, a feature data healing judgment module for preliminarily judging whether the fracture healing state conforms to the normal healing law based on the fused feature data, a time-series comparison anomaly detection module for detecting possible abnormal phenomena during the healing process by comparing the image data at different time points, a healing simulation determination module for further deeply analyzing the detected abnormal phenomena to judge whether doctor intervention is required, a bone healing progress visualization module for visually presenting the time progress and trend data of the healing, a healing intelligent advice module for providing intelligent advice for doctors based on the healing progress data and determination results, and an artificial intelligence warning module for evaluating the warning level according to the severity of the problems that occur in the healing simulation image determination.

[0007] The present invention is further configured as follows: The dynamic feature extraction module further analyzes the data after the time-series data aggregation module aggregates, corrects, and standardizes the image data of the patient at different time points, and extracts dynamic features such as bone density and healing rate; the feature fusion trend prediction module further fuses the features at multiple time points based on the dynamic features such as bone density and healing rate extracted by the dynamic feature extraction module, and generates a healing trend curve to predict the possible future development direction; the feature data healing judgment module uses the fused features and healing trend data provided by the feature fusion trend prediction module to preliminarily judge the fracture healing state and determine whether it conforms to the normal healing law; the time-series comparison anomaly detection module further detects possible abnormal phenomena during the healing process by comparing the image data at different time points after the feature data healing judgment module completes the preliminary judgment.

[0008] The present invention is further configured as follows: after the abnormal detection module for temporal sequence comparison detects an abnormality, the healing simulation determination module performs a more in-depth analysis. By simulating the healing process, it further determines whether these abnormalities require doctor intervention; the bone healing progress visualization module visually presents the determination result of the healing simulation determination module and the healing trend data, enabling the doctor to directly view the patient's healing progress and abnormal conditions; the intelligent healing advice module provides further intelligent advice to the doctor based on the progress and abnormal information presented by the bone healing progress visualization module, such as adjusting the treatment plan or suggesting further examinations; according to the advice generated by the intelligent healing advice module, if there are critical abnormalities or potential risks, the artificial intelligence warning module issues corresponding warning signals and evaluates the warning level according to the severity of the problem, reminding the doctor to take timely actions.

[0009] The present invention is further configured as follows: after the image enhancement module enhances the denoised medical image by performing operations such as contrast adjustment, brightness adjustment, and sharpening to improve the visibility of the fracture site and provide a clear image input for subsequent precise segmentation and analysis, the image denoising processing module performs denoising processing on the medical image; after the image segmentation module accurately separates the vertebral body from the surrounding tissues of the enhanced medical image through the U-Net segmentation technology to provide a precise target area for subsequent feature extraction and model construction, the image enhancement module enhances the medical image; based on the segmented fracture site image, the data augmentation module generates more samples to enrich the dataset by using data augmentation techniques such as rotation, scaling, and translation.

[0010] The present invention is further configured as follows: after the multi-scale feature fusion module combines the fracture edge and bone density feature points extracted by using the SIFT algorithm or deep learning with convolution kernels of different scales to perform multi-level fusion of global and local features and optimize the feature expression, the feature point detection module detects feature points; after the convolutional neural network construction module inputs the fused multi-scale features into a lightweight CNN architecture to build a deep learning model for fracture healing recognition and improve the recognition efficiency and accuracy, the multi-scale feature fusion module performs multi-scale feature fusion; in the CNN model constructed by the attention mechanism integration module, the convolutional neural network construction module integrates the attention mechanism to make the model pay more attention to the features of the fracture healing site.

[0011] The present invention is further configured such that: the bone healing status evaluation module uses indicators for evaluating bone density and fracture line healing in the multi-time point tracking and comparison module, combines the imaging data of the patient at different times, conducts tracking and comparison of the healing status, and forms a time series model; the multi-time point tracking and comparison module uses the healing progress analysis based on time series data in the anomaly detection module to detect abnormal conditions during the fracture healing process and automatically issue a warning to prevent treatment delays; the anomaly detection module provides warning information and diagnosis and treatment suggestions for the detected abnormal conditions in the expert-assisted treatment module to assist grass-roots doctors in further examinations or formulating treatment plans.

[0012] The present invention is further configured such that: the hardware compatibility optimization module optimizes the system hardware in the model lightweighting module to ensure compatibility and operation on various devices. At the same time, combined with the lightweight model architecture, it improves the execution efficiency and application scope of the model; the model lightweighting module, after constructing a lightweight model in the federated learning integration module, through the federated learning integration mechanism, uses the distributed data of different medical institutions to realize the joint training and update of the model and improve the generalization performance of the model; the federated learning integration module uses the integrated federated learning mechanism in the remote update and maintenance module to remotely update and maintain the model during the operation of the system to ensure the long-term application and adaptability of the system in medical institutions at all levels.

[0013] The beneficial effects of the present invention are as follows: The present invention gives play to the advantages of interdisciplinary, develops an identification system in medical imaging based on deep learning, utilizes its advantages of being able to discover detailed features and hidden rules that humans cannot detect, and then uses a large amount of structured data for automatic learning and training of the visual feature expression of abstract data to reduce the clinical missed diagnosis rate. With rigorous and scientific means for training the artificial intelligence architecture, after it has the clinical experience and knowledge of a senior attending doctor, it is popularized in local hospitals for use to achieve the role of assisting diagnosis. Furthermore, narrowing the gap in medical differentiation and making up for the deficiencies in technology and experience of local medical staff, and improving the disease diagnosis rate and reducing the missed diagnosis rate are one of the purposes of this item. Description of the Drawings

[0014] Figure 1 It is a schematic diagram of the system modules in the present invention.

[0015] Figure 2 It is a schematic diagram of the system flow of the multi-time point tracking and comparison module in the present invention. Detailed Embodiments

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the 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 of the embodiments.

[0017] Example 1 As Figure 1 shown, an image recognition system for constructing the healing of osteoporotic vertebral fractures based on artificial intelligence includes a data preprocessing module for extracting and optimizing data from original medical images, a feature extraction and model construction module for automatically extracting the features of the fracture site and constructing a suitable recognition model through deep learning technology, a bone healing monitoring and classification module for classifying the healing status of the recognized fracture images and real-time tracking the fracture healing process, and a system optimization and scalability module for improving the operation efficiency and application scope of the system to ensure its application in medical institutions at all levels; The data preprocessing module includes an image denoising processing module for removing noise or artifacts in the image, an image enhancement module for improving the visibility of the fracture site by adjusting contrast, brightness, and sharpening, an image segmentation module for separating the vertebral body from surrounding tissues using U-Net segmentation technology, and a data augmentation module for generating more samples by methods such as rotation, scaling, and translation; The feature extraction and model construction module includes a feature point detection module for extracting important feature points of fracture edges and bone density using traditional SIFT algorithms or deep learning methods, a multi-scale feature fusion module for combining global and local features of the vertebral body by introducing convolutional kernels of different scales, a convolutional neural network construction module for building a lightweight CNN architecture, and an attention mechanism integration module for concentrating more attention of the model on the fracture healing site using the attention mechanism; The bone healing monitoring and classification module includes a bone healing status evaluation module for automatically classifying the fracture healing stage based on indicators such as bone density and fracture line healing, a multi-time point tracking and comparison module for tracking the healing progress by combining image data of patients at different times and establishing a time series model, an anomaly detection module for automatically issuing a warning when an anomaly occurs during the fracture healing process, and an expert-assisted treatment module for providing decision support for grass-roots doctors and recommending further examinations or treatment plans; The system optimization and scalability module includes a hardware compatibility optimization module for optimizing the hardware requirements of the system according to the device performance of different levels of medical institutions so that it can operate in environments with limited device conditions such as county-level hospitals or township health centers, a model lightweight module for reducing the complexity and computational amount of the model through techniques such as model pruning and quantization to ensure its real-time operation on ordinary devices, a federated learning integration module for ensuring secure data sharing between different medical institutions through federated learning technology, and a remote update and maintenance module for providing the system with a remote update function to facilitate the development team to continuously optimize and upgrade the model; The image denoising processing module performs enhancement operations such as contrast, brightness, and sharpening on the denoised medical images by the image enhancement module to improve the visibility of the fracture site and provide clear image input for subsequent precise segmentation and analysis; the image enhancement module accurately separates the vertebral body from the surrounding tissues of the enhanced medical images by the U-Net segmentation technology in the image segmentation module to provide a precise target area for subsequent feature extraction and model construction; the image segmentation module generates more samples to enrich the dataset based on the segmented fracture site images by using data augmentation techniques such as rotation, scaling, and translation in the data augmentation module; The feature point detection module performs multi-level fusion of global and local features by combining the fracture edge and bone density feature points extracted by the SIFT algorithm or deep learning with convolution kernels of different scales in the multi-scale feature fusion module to optimize the feature expression; the multi-scale feature fusion module inputs the fused multi-scale features into a lightweight CNN architecture in the convolutional neural network construction module to build a deep learning model for fracture healing recognition and improve the recognition efficiency and accuracy; the convolutional neural network construction module integrates the attention mechanism in the CNN model constructed in the attention mechanism integration module to make the model pay more attention to the features of the fracture healing site; The bone healing state evaluation module uses indicators for evaluating bone density and fracture line healing conditions in the multi-time point tracking and comparison module to track and compare the healing state by combining the imaging data of the patient at different times to form a time series model; the multi-time point tracking and comparison module detects abnormal conditions during the fracture healing process based on the healing progress analysis of the time series data in the anomaly detection module and automatically issues a warning to prevent treatment delays; the anomaly detection module provides warning information and diagnosis and treatment suggestions for the detected abnormal conditions in the expert-assisted treatment module to assist grass-roots doctors in further examinations or formulating treatment plans; The hardware compatibility optimization module optimizes the system hardware in the model lightweighting module to ensure compatibility and operation on various devices. At the same time, combined with the lightweight model architecture, it improves the execution efficiency and application scope of the model; after constructing a lightweight model in the model lightweighting module, the model lightweighting module uses the federated learning integration mechanism in the federated learning integration module to realize the joint training and update of the model by using the distributed data of different medical institutions to improve the generalization performance of the model; the federated learning integration module uses the integrated federated learning mechanism in the remote update and maintenance module to remotely update and maintain the model during the operation of the system to ensure the long-term application and adaptability of the system in medical institutions at all levels.

[0018] In the above embodiments, the data preprocessing module denoises, enhances, and segments the original medical images, generates more samples through data augmentation, and optimizes the image quality. Then, the feature extraction and model construction module uses deep learning techniques to extract the features of the fracture site, constructs a lightweight model by combining multi-scale features and convolutional neural networks, and strengthens the attention to the fracture healing area through the attention mechanism. The bone healing monitoring and classification module classifies and tracks the healing process in real time by evaluating indicators such as bone density and fracture line healing, automatically detects abnormal conditions and issues warnings, and provides treatment suggestions for doctors. Finally, the system optimization and scalability module improves the operation efficiency through hardware optimization and model lightweighting, combines federated learning to ensure data sharing and remote update between different medical institutions, and guarantees the long-term application and adaptability of the system.

[0019] Among them: The U-Net segmentation technology gradually extracts the features of the image through a series of convolutional layers and pooling layers, finally obtains a feature map with a smaller resolution, and then restores the spatial resolution of the image through gradual upsampling. At the same time, by combining the features of the corresponding layers in the encoding path, the decoder not only restores the resolution, but also combines the deep feature information extracted during the encoding process to generate a more accurate segmentation result. Among them: By assigning a main direction to each key point, the SIFT algorithm can adjust the rotation of the image to ensure that the key points between rotated images can be correctly matched. It is generated based on gradient information and has good robustness to illumination changes. Due to the highly discriminative feature descriptors generated, SIFT shows high accuracy in the key point matching process. Among them: The lightweight CNN architecture reduces its computational complexity and storage requirements through various optimization methods on the premise of ensuring the accuracy and performance of the neural network model, so that the model can operate efficiently in resource-constrained environments. The purpose of building a lightweight CNN is to enable deep learning models to run in real time on ordinary devices (such as mobile devices and embedded systems) without relying on large-scale computing resources.

[0020] Embodiment 2 Such as Figure 1-2As shown in the figure, an image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence. The multi-time point tracking and comparison module includes a time series data aggregation module for aggregating and standardizing the imaging data obtained by the patient at different time points to ensure the consistency of data format and resolution, a dynamic feature extraction module for extracting dynamic features such as bone density and healing rate that change over time from the time series images, a feature fusion trend prediction module for fusing the extracted dynamic features to generate a healing trend curve to predict the possible future development direction, a feature data healing judgment module for preliminarily judging whether the fracture healing state conforms to the normal healing law based on the fused feature data, a time series comparison anomaly detection module for detecting possible abnormal phenomena during the healing process by comparing the imaging data at different time points, a healing simulation determination module for further deeply analyzing the detected abnormal phenomena to judge whether doctor intervention is required, a bone healing progress visualization module for presenting the time progress and trend data of healing in a visual way, a healing intelligent advice module for providing intelligent advice for doctors according to the healing progress data and determination results, and an artificial intelligence warning module for evaluating the warning level according to the problems in the healing simulation image determination and the severity of the problems; The dynamic feature extraction module further analyzes the data aggregated, corrected and standardized by the time series data aggregation module for the imaging data of the patient at different time points, and extracts dynamic features such as bone density and healing rate. The feature fusion trend prediction module further fuses the features at multiple time points based on the dynamic features such as bone density and healing rate extracted by the dynamic feature extraction module, and generates a healing trend curve to predict the possible future development direction. The feature data healing judgment module uses the fused features and healing trend data provided by the feature fusion trend prediction module to preliminarily judge the fracture healing state and determine whether it conforms to the normal healing law. After the feature data healing judgment module completes the preliminary judgment, the time series comparison anomaly detection module further detects possible abnormal phenomena during the healing process by comparing the imaging data at different time points; After the abnormal detection module for temporal comparison detects an anomaly, the healing simulation determination module conducts a more in-depth analysis. By simulating the healing process, it further determines whether these anomalies require doctor intervention. The bone healing progress visualization module visually presents the determination results of the healing simulation determination module and the healing trend data, enabling doctors to intuitively view the patient's healing progress and anomalies. The intelligent healing advice module provides further intelligent advice to doctors based on the progress and anomaly information presented by the bone healing progress visualization module, such as adjusting the treatment plan or suggesting further examinations. Based on the advice generated by the intelligent healing advice module, if there are critical anomalies or potential risks, the artificial intelligence warning module issues corresponding warning signals and evaluates the warning level according to the severity of the problem, reminding doctors to take timely actions.

[0021] In the above embodiment, the temporal data aggregation module aggregates and standardizes the image data at different time points to ensure data consistency. Then, the dynamic feature extraction module extracts dynamic features such as bone density and healing rate that change over time from the standardized image data, and the feature fusion trend prediction module integrates these features to generate a healing trend curve to predict the future development direction of fracture healing. Based on the trend prediction, the feature data healing judgment module makes a preliminary assessment of the healing status to determine whether it conforms to the conventional healing rules. If an anomaly is detected, the abnormal detection module for temporal comparison further identifies potential abnormal phenomena during the healing process by comparing the image data at different time points. Subsequently, the healing simulation determination module conducts an in-depth analysis of these anomalies to determine whether doctor intervention is required. The system also visually displays the healing trend and anomalies through the bone healing progress visualization module to help doctors intuitively view the patient's healing status. Based on this, the intelligent healing advice module provides intelligent advice for doctors to adjust the treatment plan or conduct further examinations. If there are serious problems, the artificial intelligence warning module issues a warning to remind doctors to handle them in a timely manner.

[0022] Working principle: When the present invention is in use, the image denoising processing module processes medical images to remove noise or artifacts, thereby obtaining clearer images. Next, the image enhancement module adjusts the contrast, brightness, sharpness, etc. on the denoised images to improve the visibility of the fracture site and ensure the clarity of the fracture area.

[0023] Then, the image segmentation module uses U-Net segmentation technology to separate the vertebral body from the surrounding tissues and extract the precise area of the fracture. To address the issue of insufficient data, the data augmentation module generates more samples through rotation, scaling, translation, etc., enriching the dataset and enhancing the generalization ability of the model. After the data preprocessing is completed, it enters the feature extraction and model construction stage. The feature point detection module extracts key feature points such as fracture edges and bone density using the SIFT algorithm or deep learning methods.

[0024] Next, the multi-scale feature fusion module uses convolutional kernels of different scales to globally and locally fuse these features to optimize the feature representation and enhance the understanding of the fracture area. The fused features are input into the convolutional neural network construction module, which constructs a lightweight CNN architecture to ensure that the model can operate efficiently with limited device resources. Finally, the attention mechanism integration module applies the attention mechanism to the constructed CNN, enabling the model to focus more on the details of the fracture healing area and further improving the recognition accuracy. After the feature extraction is completed, the system enters the healing monitoring and classification stage. The bone healing status evaluation module classifies the healing status based on indicators such as bone density and fracture line healing. Then, the multi-time point tracking and comparison module combines the imaging data of the patient at different time points to construct a time series model to track the fracture healing progress in real time. The anomaly detection module monitors the healing progress, detects whether there are abnormal situations, and automatically issues warnings.

[0025] Finally, the expert-assisted treatment module provides intelligent diagnosis and treatment suggestions for primary care physicians. The entire system ensures its adaptability to different medical scenarios through the system optimization and scalability module, including the hardware compatibility optimization module and the model lightweight module, enabling the system to operate in an environment with limited device resources, and realizing data sharing between institutions and remote updating of the model through the federated learning integration module and the remote update and maintenance module to ensure the long-term application of the system.

[0026] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence, characterized in that: It includes a data preprocessing module for extracting and optimizing data from raw medical images, a feature extraction and model building module for automatically extracting features of fracture sites and building a suitable recognition model through deep learning technology, a bone healing monitoring and classification module for classifying the healing status of identified fracture images and tracking the fracture healing process in real time, and a system optimization and scalability module for improving the operating efficiency and scope of the system to ensure that it can be used in medical institutions at all levels; The data preprocessing module includes an image denoising module for removing noise or artifacts in the image, an image enhancement module for improving the visibility of the fracture site by adjusting contrast, brightness, and sharpening, an image segmentation module for separating the vertebral body from the surrounding tissue using U-Net segmentation technology, and a data expansion module for generating more samples by methods such as rotation, scaling, and translation; The feature extraction and model building module includes a feature point detection module for extracting fracture edges and important feature points of bone density using traditional SIFT algorithm or deep learning method, a multi-scale feature fusion module for combining global and local features of vertebral bodies by introducing convolution kernels of different scales, a convolutional neural network building module for building a lightweight CNN architecture, and an attention mechanism integration module for using the attention mechanism to focus the model's attention more on the fracture healing site.

2. According to claim 1, an image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence is characterized in that: The bone healing monitoring and classification module includes a bone healing status assessment module for automatically classifying the fracture healing stage based on indicators such as bone density and fracture line healing status, a multi-time point tracking and comparison module for tracking the healing progress and establishing a time series model by combining imaging data of patients at different stages, an abnormality detection module for automatically issuing an early warning when abnormalities are detected during the fracture healing process, and an expert auxiliary treatment module for providing decision support to primary care doctors and recommending further examinations or treatment plans; The system optimization and scalability module includes a hardware compatibility optimization module for optimizing the hardware requirements of the system according to the equipment performance of medical institutions at different levels so that it can operate in environments with limited equipment conditions such as county hospitals or township health centers; a model lightweight module for reducing the complexity and computational complexity of the model through model pruning, quantization and other technologies to ensure that it can run in real time on ordinary devices; a federated learning integration module for ensuring the secure sharing of data between different medical institutions through federated learning technology; and a remote update and maintenance module for providing the system with remote update functions to facilitate the development team to continuously optimize and upgrade the model.

3. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 2, characterized in that: The multi-time point tracking and comparison module includes a time series data aggregation module for aggregating and standardizing the image data obtained from the patient at different time points to ensure consistency of data format and resolution, a dynamic feature extraction module for extracting dynamic features such as bone density and healing rate that change with time from time series images, a feature fusion trend prediction module for fusing the extracted dynamic features to generate a healing trend curve to predict possible future development directions, a feature data healing judgment module for making a preliminary judgment on whether the fracture healing state conforms to the normal healing law based on the fused feature data, a time series comparison anomaly detection module for detecting possible abnormal phenomena in the healing process by comparing image data at different time points, a healing simulation judgment module for further in-depth analysis of the detected abnormal phenomena to determine whether a doctor's intervention is required, a bone healing progress visualization module for presenting the time progress and trend data of healing in a visual manner, a healing intelligent suggestion module for providing intelligent suggestions to doctors based on the healing progress data and judgment results, and an artificial intelligence early warning module for judging problems in the healing simulation images and evaluating the early warning level according to the severity of the problems.

4. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 3, characterized in that: The dynamic feature extraction module further analyzes the data obtained by aggregating, correcting and standardizing the imaging data of the patient at different time points by the time series data aggregation module, and extracts dynamic features such as bone density and healing rate; the feature fusion trend prediction module further fuses the features of multiple time points based on the dynamic features such as bone density and healing rate extracted by the dynamic feature extraction module, and generates a healing trend curve to predict possible future development directions; the feature data healing judgment module uses the fusion features and healing trend data provided by the feature fusion trend prediction module to preliminarily judge the fracture healing state and whether it conforms to normal healing rules; after the feature data healing judgment module completes the preliminary judgment, the time series comparison anomaly detection module further detects abnormal phenomena that may occur in the healing process by comparing the imaging data at different time points.

5. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 4, characterized in that: After the timing comparison anomaly detection module detects an anomaly, the healing simulation judgment module conducts a more in-depth analysis and further determines whether these anomalies require doctor intervention by simulating the healing process; the bone healing progress visualization module visualizes the judgment results of the healing simulation judgment module and the healing trend data, so that the doctor can intuitively view the patient's healing progress and abnormal conditions; the healing intelligent suggestion module provides further intelligent suggestions to the doctor based on the progress and abnormal information displayed by the bone healing progress visualization module, such as adjusting the treatment plan or suggesting further examinations; the artificial intelligence early warning module issues corresponding early warning signals based on the suggestions generated by the healing intelligent suggestion module if there are key anomalies or potential risks, and rates the early warning level according to the severity of the problem, reminding the doctor to take timely action.

6. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 1, characterized in that: The image denoising module performs contrast, brightness, sharpening and other enhancement operations on the denoised medical image in the image enhancement module to improve the visibility of the fracture site and provide clear image input for subsequent precise segmentation and analysis; the image enhancement module uses the U-Net segmentation technology to accurately separate the vertebral body from the surrounding tissues in the image segmentation module to the enhanced medical image, and provides a precise target area for subsequent feature extraction and model construction; the image segmentation module generates more samples in the data expansion module based on the segmented fracture site image using data enhancement technologies such as rotation, scaling, and translation to enrich the data set.

7. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 1, characterized in that: The feature point detection module uses the SIFT algorithm or deep learning to extract the fracture edge and bone density feature points in the multi-scale feature fusion module, combined with convolution kernels of different scales, to perform multi-level fusion of global and local features and optimize feature expression; the multi-scale feature fusion module inputs the fused multi-scale features into the lightweight CNN architecture in the convolutional neural network construction module, builds a deep learning model for fracture healing identification, and improves identification efficiency and accuracy; the convolutional neural network construction module integrates the attention mechanism in the CNN model constructed by the attention mechanism integration module, so that the model pays more attention to the characteristics of the fracture healing site.

8. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 2, characterized in that: The bone healing status assessment module uses indicators for assessing bone density and fracture line healing in the multi-time point tracking and comparison module, combined with imaging data of patients at different periods, to track and compare the healing status to form a time series model; the multi-time point tracking and comparison module analyzes the healing progress based on time series data in the abnormality detection module, detects abnormalities in the fracture healing process, and automatically issues an early warning to prevent delayed treatment; the abnormality detection module provides early warning information and diagnosis and treatment suggestions for detected abnormalities in the expert auxiliary treatment module, and assists primary doctors in conducting further examinations or formulating treatment plans.

9. The image recognition system for osteoporotic vertebral fracture healing based on artificial intelligence according to claim 2, characterized in that: The hardware compatibility optimization module optimizes the system hardware in the model lightweight module to ensure compatible operation on multiple devices, and combines the lightweight model architecture to improve the execution efficiency and scope of application of the model; after the lightweight model is constructed in the federated learning integration module, the model lightweight module uses the federated learning integration mechanism to utilize the distributed data of different medical institutions to achieve joint training and updating of the model, thereby improving the generalization performance of the model; the federated learning integration module utilizes the integrated federated learning mechanism in the remote update and maintenance module to remotely update and maintain the model during system operation, thereby ensuring the long-term application and adaptability of the system in medical institutions at all levels.

Images