AI-based intelligent classification management system for image and video acquisition data

Abstract

This invention discloses an intelligent classification and management system for image and video acquisition data based on artificial intelligence, belonging to the field of medical image processing technology. It includes a medical image acquisition module, comprising an image acquisition unit and a preprocessing unit. The image acquisition unit acquires medical image data; the preprocessing unit responds to the image acquisition unit and performs preprocessing on the acquired medical image data, including noise reduction, contrast enhancement, and format conversion; a feature extraction module extracts features from the preprocessed medical image data, including extracting feature vectors from the medical image data. This invention improves the classification efficiency and accuracy of medical image data through automated feature extraction, intelligent classification, dynamic optimization, and data fusion management, enhancing data management capabilities and providing strong support for disease diagnosis and prognosis.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to an intelligent classification and management system for image and video acquisition data based on artificial intelligence. Background Technology

[0002] With the rapid development of information technology, the amount of image and video data collected has exploded. This data is widely used in fields such as security monitoring, intelligent transportation, medical imaging, and entertainment, and is also classified and managed in the corresponding fields.

[0003] Regarding this research, application CN202311389989.5 provides an artificial intelligence-based image classification system. This technical solution includes a first image acquisition unit for acquiring uploaded mammogram images to be classified; a first image preprocessing unit for dividing the mammogram images to be classified into target regions, obtaining target region images corresponding to the mammogram images; and an artificial intelligence model unit for inputting the mammogram images to be classified and the target region images into a trained artificial intelligence model to obtain the classification result of the mammogram images to be classified. This technical solution solves the technical problem of low classification accuracy when classifying mammogram images using high-level features extracted from mammogram images through convolutional neural networks.

[0004] Another application, CN202011158352.1, provides an artificial intelligence-based medical image classification and processing system. This technical solution acquires image information of the patient's lesion using a camera device, tags the uploaded image data, and uploads it to a terminal server to expand and share the data. Simultaneously, it compares the patient's image information with image information stored in the database, matching similar case images. This allows for accurate analysis of the patient's medical image data and the generation of medical examination reports, thereby assisting doctors in improving the efficiency and accuracy of medical examinations.

[0005] However, the above-mentioned technical solutions ignore the noise, artifacts and other uncertainties that may exist in medical images. These factors can affect doctors' annotation of the patient's disease type, lesion location and the texture, shape and density associated with the disease type and lesion location, leading to errors in the annotation of medical images. This makes it difficult to classify and retrieve massive amounts of medical images, thus affecting the efficiency of clinical diagnosis. Summary of the Invention

[0006] In view of the problems existing in the field of medical image processing technology, the present invention is proposed.

[0007] Therefore, one of the objectives of this invention is to provide an intelligent classification and management system for image and video acquisition data based on artificial intelligence. Through functions such as automated feature extraction, intelligent classification, dynamic optimization, and data fusion management, it improves the classification efficiency and accuracy of medical image data, enhances data management capabilities, and provides strong support for disease diagnosis and prognosis.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] This invention provides an intelligent classification and management system for image and video acquisition data based on artificial intelligence, comprising:

[0010] A medical image acquisition module, comprising an image acquisition unit and a preprocessing unit;

[0011] The image acquisition unit is used to acquire medical image data;

[0012] The preprocessing unit responds to the image acquisition unit and is used to preprocess the acquired medical image data. The preprocessing of the medical image data includes noise reduction, contrast enhancement, and format conversion.

[0013] The feature extraction module is used to extract features from the preprocessed medical image data. The feature extraction includes extracting feature vectors from the medical image data. The feature vectors include texture features, shape features, and grayscale features of the image in the medical image data.

[0014] A data processing module, which responds to the feature vector, is used to construct a database based on the extracted feature vector, classify the medical image data in the database according to different feature vectors, and generate a classification model;

[0015] A data fusion management module, comprising a training unit, a classification unit, and a learning unit;

[0016] The training unit is used to train the classification model based on different feature vectors. The training method includes annotating medical image data in the classification model. The information annotation includes annotating the disease type and lesion location of the medical image data.

[0017] The classification unit is used to input the extracted feature vectors into the trained classification model to automatically classify medical image data.

[0018] The learning unit is used to optimize the classification model based on feedback from doctors and newly acquired medical image data.

[0019] In a preferred embodiment of the present invention, in the learning unit, the doctor's feedback includes the accuracy of the classification results, wherein the accuracy of the classification results includes correctness feedback and misclassification feedback.

[0020] The correctness feedback indicates to the doctor whether the classification result of the classification model is correct;

[0021] The misclassification feedback indicates to the doctor the misclassification situation of the classification model;

[0022] The doctor's feedback also includes the identification of key areas, which includes feedback on lesion areas and the importance of features;

[0023] The feedback on the lesion area indicates to the doctor whether the classification model has accurately identified the lesion area;

[0024] The feature importance feedback indicates to the doctor the key and redundant features of the medical image data in the classification model. The key features include texture, shape and density.

[0025] In a preferred embodiment of the present invention, the data processing module classifies the medical image data according to different feature vectors, including dividing the medical image data into... , ,..., ,in, The nth category of medical image data is defined. Differences in the feature vectors among different categories of medical image data are obtained. Feature vectors are extracted from newly acquired medical image data in future time periods. The feature vectors are compared with the feature vectors in the already classified medical image data. Based on the comparison results, the corresponding feature vectors are input into the classification model to classify the acquired medical image data. When medical image data is retrieved in the classification model in future time periods, the feature vectors are input into the classification model to obtain the medical image data corresponding to the feature vectors.

[0026] In a preferred embodiment of the present invention, the feature vectors of medical image data are distinguished in the classification model. The distinction is made by classifying the feature vectors into early-stage feature vectors, mid-stage feature vectors, and late-stage feature vectors according to the acquisition time of the medical image data. The feature change from the early-stage feature vector to the mid-stage feature vector is marked as a stage θ1 change, and the feature change from the mid-stage feature vector to the late-stage feature vector is marked as a stage θ2 change. The disease type, lesion location, and symptoms corresponding to the early-stage feature vector are obtained in the classification model. In newly acquired medical image data in the future, if the feature vector corresponding to the medical image data matches the disease type, lesion location, and symptoms in the early-stage feature vector, the system determines that the feature vector of the newly acquired medical image data will change towards the mid-stage feature vector; otherwise, no determination is made.

[0027] In a preferred embodiment of the present invention, the feature vectors of medical image data are distinguished in the classification model. The distinction further includes classifying the feature vectors into mild, moderate, and severe feature vectors according to the degree of lesion. The feature change from mild to moderate feature vectors is marked as a δ1 stage change, and the feature change from moderate to late-stage feature vectors is marked as a δ2 stage change. The disease type, lesion location, and symptoms corresponding to the mild feature vectors are obtained in the classification model. In newly acquired medical image data in future periods, if the feature vector corresponding to the medical image data matches the disease type, lesion location, and symptoms in the mild feature vector, the system determines that the feature vector of the newly acquired medical image data will change towards the moderate feature vector; otherwise, no determination is made.

[0028] In a preferred embodiment of the present invention, if the feature vector of newly acquired medical image data in the future period does not change toward the intermediate feature vector or the moderate feature vector, then the newly acquired medical image data is reclassified in the classification model; otherwise, no classification is performed.

[0029] In a preferred embodiment of the present invention, when the system determines that the feature vector of newly acquired medical image data will change towards the intermediate feature vector, the degree of change of the early feature vector towards the intermediate feature vector at different time points is calculated according to the following formula:

[0030] ;in, Represents the early-stage feature vector;

[0031] In the formula, Represents the mid-term eigenvector. Represents the change vector of the eigenvector;

[0032] The magnitude of change of the feature vector is calculated based on the change vector. The magnitude of change of the feature vector measures the magnitude of the change of the feature vector at different time points, as follows:

[0033] ;

[0034] In the formula, The Euclidean norm represents the change in the eigenvector and is used to quantify the magnitude of the change.

[0035] In a preferred embodiment of the present invention, the rate of change of the feature vector is calculated based on the magnitude of the change, wherein the rate of change is the speed at which the feature vector changes per unit time, as follows:

[0036] ;

[0037] In the formula, The time interval is represented by r, which represents the rate of change of the eigenvector.

[0038] In a preferred embodiment of the present invention, a change threshold is preset based on the degree of change of the calculated early feature vector to the mid-term feature vector at different time points, and a first time point, a second time point, and a third time point are preset within the period based on the change threshold. At any time point, if the degree of change of the early feature vector to the mid-term feature vector exceeds the change threshold, the system performs a new classification on the medical image data corresponding to the change threshold; otherwise, no classification is performed. Beneficial effects

[0039] 1. The feature extraction module automatically extracts the texture, shape, and grayscale features of medical images, reducing the workload of manual annotation and improving the efficiency of data processing. The classification model automatically classifies the extracted feature vectors, enabling rapid and accurate classification of large amounts of medical image data and improving diagnostic efficiency.

[0040] 2. The learning unit optimizes the classification model based on doctors' feedback (such as the accuracy of classification results, key area identification, etc.) and newly collected medical image data, so that the model can continuously adapt to new data and clinical needs, and further improve the accuracy of classification.

[0041] 3. The data processing module classifies medical image data based on feature vectors and builds a database, which facilitates doctors to quickly retrieve and access specific types of image data, improving the convenience and efficiency of data management. Furthermore, by comparing feature vectors, newly acquired medical image data can be quickly classified, ensuring the orderly storage and management of data.

[0042] 4. The system can dynamically adjust the classification criteria according to preset change thresholds, ensuring timely reclassification of image data during disease progression, thus improving the system's flexibility and adaptability. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0044] Figure 1 This is a schematic diagram of the modular structure of the AI-based intelligent classification management system for image and video acquisition data according to an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the process structure of an embodiment of the present invention;

[0046] The diagram is labeled as follows: 110 - Medical image acquisition module; 1101 - Image acquisition unit; 1102 - Preprocessing unit; 120 - Feature extraction module; 130 - Data processing module; 140 - Data fusion management module; 1401 - Training unit; 1402 - Classification unit; 1403 - Learning unit. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0048] Because existing technologies ignore noise, artifacts, and other uncertainties that may exist in medical images, these factors can affect doctors' annotation of patient disease types, lesion locations, and the textures, shapes, and densities associated with disease types and lesion locations. This can lead to errors in the annotation of medical images, making it difficult to classify and retrieve massive amounts of medical images, thus affecting the efficiency of clinical diagnosis.

[0049] Based on this, the present invention proposes an intelligent classification and management system for image and video acquisition data based on artificial intelligence. Through functions such as automated feature extraction, intelligent classification, dynamic optimization, and data fusion management, it improves the classification efficiency and accuracy of medical image data, enhances data management capabilities, and provides strong support for disease diagnosis and prognosis.

[0050] The present solution will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0051] Reference Figures 1 to 2 As one embodiment of the present invention, this embodiment provides an intelligent classification and management system for image and video acquisition data based on artificial intelligence, including:

[0052] The medical image acquisition module 110 includes an image acquisition unit 1101 and a preprocessing unit 1102.

[0053] The image acquisition unit 1101 is used to acquire medical image data;

[0054] The preprocessing unit 1102 responds to the image acquisition unit and is used to preprocess the acquired medical image data. The preprocessing of the medical image data includes noise reduction, contrast enhancement and format conversion.

[0055] The feature extraction module 120 is used to extract features from the preprocessed medical image data. Feature extraction includes extracting feature vectors from the medical image data. The feature vectors include texture features, shape features, and grayscale features of the image in the medical image data.

[0056] The data processing module 130 responds to the feature vector, which is used to construct a database based on the extracted feature vector, classify medical image data in the database according to different feature vectors, and generate a classification model.

[0057] The data fusion management module 140 includes a training unit 1401, a classification unit 1402, and a learning unit 1403.

[0058] The training unit 1401 is used to train the classification model based on different feature vectors. The training method includes annotating medical image data in the classification model. The information annotation includes the annotation of disease type and lesion location in the medical image data.

[0059] The classification unit 1402 is used to input the extracted feature vectors into the trained classification model to automatically classify medical image data;

[0060] In this embodiment, the classification result can be a specific disease type or a disease probability distribution;

[0061] Learning unit 1403 is used to optimize the classification model based on doctor feedback and newly acquired medical image data;

[0062] In this embodiment, a complete AI-based intelligent classification and management system for medical image data was constructed, covering the entire process from image acquisition, preprocessing, feature extraction to data processing and model optimization, which can meet the needs of medical image data from acquisition to classification management.

[0063] In the learning unit, the doctor's feedback includes the accuracy of the classification results, which includes feedback on correctness and feedback on misclassification.

[0064] Correctness feedback indicates to doctors whether the classification model's classification results are correct;

[0065] For example, the model might classify an image as "normal," but a doctor might diagnose that the image actually contains a lesion.

[0066] Misclassification feedback indicates to doctors instances of misclassification by the classification model;

[0067] For example, misclassifying a benign lesion as malignant, or vice versa, such feedback is particularly important for adjusting the decision boundary of the classification model;

[0068] The doctor's feedback also includes the identification of key areas, which includes feedback on the lesion area and the importance of features;

[0069] The feedback on the lesion area indicates to doctors whether the classification model has accurately identified the lesion area;

[0070] For example, in lung CT images, does the model accurately identify the location and extent of lung nodules?

[0071] Feature importance feedback helps doctors identify key and redundant features of medical image data in classification models. Key features include texture, shape, and density.

[0072] In this embodiment, the doctor's professional feedback is incorporated into the optimization process of the classification model, so that the model can be adjusted in combination with clinical experience and actual needs.

[0073] Doctors can provide feedback on the correctness of the classification results and the occurrence of misclassifications, which helps to correct model errors in a timely manner and improve classification accuracy.

[0074] Furthermore, the doctor pointed out the importance of the lesion area and key features, enabling the model to more accurately identify and focus on imaging features that are important for diagnosis, avoiding interference from redundant information;

[0075] In the data processing module, medical image data is classified according to different feature vectors, including dividing the medical image data into... , ,..., ,in, This represents the nth class of medical image data. Differences in feature vectors are obtained from different classes of medical image data. Feature vectors are extracted from newly acquired medical image data in future time periods. The feature vectors are compared with the feature vectors in the already classified medical image data. Based on the comparison results, the corresponding feature vectors are input into the classification model to classify the acquired medical image data. When medical image data is retrieved in the classification model in future time periods, the feature vectors are input into the classification model to obtain the medical image data corresponding to the feature vectors.

[0076] This embodiment clarifies the specific method for classifying medical image data based on feature vectors, as well as the processing method for newly acquired data in the future. By dividing medical image data into different categories and obtaining the feature vector differences between different categories, the data classification becomes clearer and easier to manage and retrieve.

[0077] By extracting and comparing feature vectors from newly acquired medical image data in the future, it is possible to classify them into appropriate categories in a timely manner, ensuring the timeliness and accuracy of the classification model.

[0078] During retrieval, the corresponding medical image data can be obtained by inputting the feature vector, which improves retrieval efficiency and makes it easier for doctors to quickly access the image data they need.

[0079] This embodiment emphasizes that the feature vectors of medical image data are distinguished in the classification model. The distinction is made by classifying the feature vectors into early-stage, mid-stage, and late-stage feature vectors according to the acquisition time of the medical image data. Specifically, the feature change from the early-stage feature vector to the mid-stage feature vector is marked as stage θ1, and the feature change from the mid-stage feature vector to the late-stage feature vector is marked as stage θ2. The classification model then obtains the disease type, lesion location, and symptoms corresponding to the early-stage feature vectors. In newly acquired medical image data in future time periods, if the feature vectors corresponding to the medical image data match the disease type, lesion location, and symptoms in the early-stage feature vectors, the system determines that the feature vectors of the newly acquired medical image data will change towards the mid-stage feature vectors; otherwise, no determination is made.

[0080] In this embodiment, by introducing a time dimension to distinguish between the early, middle and late stages of the feature vector, the disease development process can be better reflected.

[0081] By marking changes in the disease stages, we can clearly see the evolution of the disease from the early stage to the middle stage and then to the late stage, providing doctors with dynamic information about the disease.

[0082] Furthermore, the system can predict whether newly acquired image data will change towards the intermediate feature vector based on the disease type, lesion location, and symptoms corresponding to the previous feature vector, thereby making an early judgment on the development trend of the disease, which is helpful for early intervention and treatment.

[0083] This distinction is of significant clinical importance for subsequently judging the patient's disease progression;

[0084] Early feature vectors are used to identify early disease manifestations that may be mild or atypical. For example, in early lung cancer, CT images may only show small nodules. Feature vectors mainly reflect the small size of the nodules and the clarity of their edges.

[0085] Intermediate-stage eigenvectors reflect the gradual emergence of imaging manifestations as the disease progresses. For example, in intermediate-stage lung cancer, CT images may show nodule enlargement, irregular margins, and ground-glass opacities, and the eigenvectors will reflect these changes.

[0086] Late-stage feature vectors are formed when the imaging findings are more severe in advanced stages of the disease. For example, in advanced lung cancer, CT images may show large tumors, lymph node metastases, pleural effusions, etc., and the feature vectors will contain these complex features.

[0087] Based on the above, the feature vectors of medical image data are differentiated in the classification model. This differentiation includes classifying feature vectors into mild, moderate, and severe feature vectors according to the severity of the lesion. Specifically, the feature change from mild to moderate feature vectors is marked as stage δ1, and the feature change from moderate to severe feature vectors is marked as stage δ2. The classification model then obtains the disease type, lesion location, and symptoms corresponding to the mild feature vectors. In newly acquired medical image data in future periods, if the feature vectors corresponding to the medical image data match the disease type, lesion location, and symptoms of the mild feature vectors, the system determines that the feature vectors of the newly acquired medical image data will change towards the moderate feature vectors; otherwise, no determination is made.

[0088] In this embodiment, the classification dimensions of the feature vectors are further refined, and the feature vectors are distinguished from the perspective of the degree of lesion, which enriches the description of the disease state.

[0089] It helps to classify and manage patients with different disease severity, improve the efficiency of medical resource utilization, and optimize disease management processes;

[0090] Mild eigenvectors are used when the lesion is mild, and the imaging features are relatively simple. For example, in early cerebral infarction, MRI images may only show mild local edema of brain tissue, and the eigenvectors mainly reflect the extent and density of the edematous area;

[0091] In moderate eigenvectors, the imaging features become more pronounced as the lesion worsens. For example, in intermediate-stage cerebral infarction, MRI images may show extensive cerebral edema and localized brain atrophy, and the eigenvectors will reflect these changes.

[0092] Severe feature vectors indicate that when the lesion is severe, the imaging features are complex and diverse. For example, in late-stage cerebral infarction, MRI images may show large areas of brain tissue necrosis, ventricular deformation, and widening of sulci, and the feature vector will contain these complex features.

[0093] Distinguishing feature vectors has advantages in judging disease trends, including dynamically monitoring disease progression, predicting disease prognosis, assisting clinical decision-making, and improving diagnostic accuracy.

[0094] Furthermore, based on the system's determination, if the feature vector of newly acquired medical image data in the future period does not change towards the intermediate feature vector or the moderate feature vector, then the newly acquired medical image data will be reclassified in the classification model; otherwise, no new classification will be performed.

[0095] In this embodiment, the handling method when newly acquired medical image data does not conform to the expected trend of change is clarified, ensuring the accuracy and stability of the classification model;

[0096] If the newly collected feature vectors do not change toward the expected intermediate or moderate feature vectors, the data is reclassified to avoid classification errors caused by misjudgment and ensure the reliability of the classification results.

[0097] This mechanism enables the classification model to correct the impact of outlier data in a timely manner, maintain the stability and adaptability of the model, and avoid the degradation of model performance due to interference from individual data.

[0098] Based on the above, when the system determines that the feature vector of newly acquired medical image data will change towards the intermediate feature vector, the degree of change during the period of change from the early feature vector to the intermediate feature vector at different time points is calculated using the following formula:

[0099] ;in, Represents the early-stage feature vector;

[0100] In the formula, Represents the mid-term eigenvector. Represents the change vector of the eigenvector;

[0101] The magnitude of change of the feature vector is calculated based on the change vector. The magnitude of change of the feature vector measures the magnitude of the change of the feature vector at different time points, as follows:

[0102] ;

[0103] In the formula, The Euclidean norm represents the change in the eigenvector and is used to quantify the magnitude of the change.

[0104] In a preferred embodiment of the present invention, the rate of change of the feature vector is calculated based on the magnitude of the change, wherein the rate of change is the speed at which the feature vector changes per unit time, as follows:

[0105] ;

[0106] In the formula, The time interval is represented by r, which represents the rate of change of the eigenvector.

[0107] In this embodiment, the rate of change of the feature vector is further quantified, providing a more accurate indicator for assessing the rate of disease progression.

[0108] Based on the degree of change of the calculated early feature vector to the mid-term feature vector at different time points, a change threshold is preset, and based on the change threshold, a first time point, a second time point, and a third time point are preset within the period. At any time point, if the degree of change of the early feature vector to the mid-term feature vector exceeds the change threshold, the system will perform a new classification on the medical image data corresponding to the change threshold; otherwise, no classification will be performed.

[0109] In this embodiment, by introducing a change threshold and a preset time point, dynamic monitoring and classification adjustment of the degree of change in feature vectors are achieved;

[0110] By setting preset change thresholds and time points, the system can monitor changes in feature vectors in real time. Once the degree of change exceeds the threshold, it will immediately issue an early warning and perform a new classification, thereby improving the sensitivity and response speed to disease changes.

[0111] In summary, this application improves the efficiency and accuracy of medical image data classification and enhances data management capabilities through functions such as automated feature extraction, intelligent classification, dynamic optimization, and data fusion management, providing strong support for disease diagnosis and prognosis.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An image and video acquisition data intelligent classification management system based on artificial intelligence, characterized in that, The system comprises a medical image acquisition module, a feature extraction module, a data processing module, and a data fusion management module, wherein the data fusion management module comprises a training unit, a classification unit, and a learning unit. The training unit is configured to train a classification model according to different feature vectors. The classification unit is configured to input the extracted feature vectors into the trained classification model to automatically classify medical image data. The learning unit is configured to optimize the classification model according to the feedback of doctors and newly acquired medical image data, wherein the feedback of doctors includes the accuracy of classification results, which includes correctness feedback and misclassification feedback. The feedback of doctors also includes the identification of key regions, which includes lesion region feedback and feature importance feedback. The lesion region feedback indicates to doctors whether the classification model accurately identifies the lesion region. The feature importance feedback indicates to doctors the key features and redundant features of medical image data in the classification model, wherein the key features include texture, shape, and density. In the classification model, the feature vectors of medical image data are distinguished, including distinguishing the feature vectors into early feature vectors, mid-term feature vectors, and late feature vectors according to the acquisition time of medical image data; wherein the feature change from the early feature vectors to the mid-term feature vectors is marked as θ1 stage change, and the feature change from the mid-term feature vectors to the late feature vectors is marked as θ2 stage change; and the disease type, lesion site, and symptoms corresponding to the early feature vectors are obtained in the classification model, and in the future period, if the feature vectors corresponding to the newly acquired medical image data meet the disease type, lesion site, and symptoms in the early feature vectors, the system determines that the feature vectors of the newly acquired medical image data will change towards the mid-term feature vectors; otherwise, it is not determined. According to the result of system determination, if the feature vectors of the newly acquired medical image data in the future period do not change towards the mid-term feature vectors, the newly acquired medical image data is classified in the classification model; otherwise, it is not classified. A change threshold is preset according to the change degree of the early feature vectors towards the mid-term feature vectors at different time points during the period, and a first time point, a second time point, and a third time point are preset within the period based on the change threshold, wherein if the change degree of the early feature vectors towards the mid-term feature vectors exceeds the change threshold at any time point, the system classifies the medical image data corresponding to the change threshold; otherwise, it is not classified. The correctness feedback indicates to doctors whether the classification results of the classification model are correct, and the misclassification feedback indicates to doctors the misclassification of the classification model. 2.The artificial intelligence-based image and video acquisition data intelligent classification management system of claim 1, wherein, ​ 3.The artificial intelligence-based image and video acquisition data intelligent classification management system of claim 1, wherein, In the data processing module, the medical image data is classified according to different feature vectors, including dividing the medical image data into wherein, indicates the n-th type of divided medical image data, the differences in the feature vectors are obtained in different types of medical image data, and the feature vectors of newly collected medical image data in a future period are extracted, the feature vectors are compared with the feature vectors in the classified medical image data, and the corresponding feature vectors are input into the classification model according to the comparison results, so as to classify the collected medical image data. When searching for medical image data in the classification model in the future period, the feature vectors are input into the classification model to obtain the medical image data corresponding to the feature vectors. 4.The artificial intelligence-based image and video acquisition data intelligent classification management system of claim 1, wherein, The feature vector of the medical image data is distinguished in the classification model, and the distinguishing manner further includes distinguishing the feature vector into a mild feature vector, a moderate feature vector and a severe feature vector according to lesion degrees; wherein a feature change of the mild feature vector to the moderate feature vector is marked as a δ1 stage change, and a feature change of the moderate feature vector to the severe feature vector is marked as a δ2 stage change; and the disease type, the lesion site and the symptom corresponding to the mild feature vector are obtained in the classification model, and in newly collected medical image data in a future period, if a feature vector corresponding to the medical image data is consistent with the disease type, the lesion site and the symptom in the mild feature vector, then the system determines that the feature vector of the newly collected medical image data will change to the moderate feature vector; otherwise, it is not determined. 5.The artificial intelligence-based image and video acquisition data intelligent classification management system of claim 1, wherein, When the system determines that the feature vector of the newly collected medical image data will change to the moderate feature vector, then the change degree of the feature vector of the medical image data at different time points in the period from the previous feature vector to the moderate feature vector is calculated, and the change degree is calculated according to the following formula: ； wherein v 前期 represents the early stage feature vector; In the formula, v 中期 a medium-term feature vector, a change vector of the feature vector; The change amplitude of the feature vector is calculated based on the change vector, and the change amplitude of the feature vector is a measure of the change size of the feature vector at different time points, and is calculated according to the following formula: ； In the formula, Euclidean norm representing the change of the feature vector, for quantifying the change size. 6.The artificial intelligence-based image and video acquisition data intelligent classification management system of claim 5, wherein, The change rate of the feature vector is calculated according to the change size, and the change rate is the change speed of the feature vector in unit time, and is calculated according to the following formula: ； In the formula, denotes the time interval, and r denotes the rate of change of the feature vector.

Images