Artificial intelligence-based nervous system medical image case database construction system

Abstract

The invention relates to the technical field of medical images, in particular to a nervous system medical image case database construction system based on artificial intelligence, which comprises a nervous system medical image database module, a teaching interactive learning platform module, a user self-test module and a user online examination module. The teaching system truly restores a diagnosis report written in a clinical working environment of a doctor, organically combines book knowledge with clinical practice, can arouse the learning enthusiasm and initiative of imaging medical students and living training users aiming at complex nervous system medical imaging teaching, improves the teaching quality and practical skills, and improves the teaching efficiency. The learning effect is fed back in real time through the computer, rapid fusion of theoretical knowledge and clinical practice is promoted, accurate lesson preparation of teachers is facilitated, a flexible and convenient education mode and an accurate and individualized training mode are achieved, the training process of image resident physicians is accelerated, the teaching efficiency and quality are improved, and the system adapts to a modern teaching mode.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a system for building a nervous system medical imaging case database based on artificial intelligence. Background Art

[0002] Medical imaging is a science that studies the interaction between a certain medium (such as X-rays, electromagnetic fields, ultrasound, etc.) and the human body, and expresses the structure and density of internal tissues and organs in the form of images, so that diagnostic physicians can make judgments based on the information provided by the images, thereby evaluating the health status of the human body. Medical imaging can be used as a medical auxiliary means for diagnosis and treatment, and can also be used as a scientific research means in life science research. Diagnosis mainly includes fluoroscopy, radiographs, CT, MRI, ultrasound, digital subtraction, angiography, etc., and treatment is mainly used in interventional therapy, radiotherapy, etc. The diagnosis and treatment of many diseases are highly dependent on medical imaging, especially central nervous system diseases, cardiovascular diseases, etc. Among them, the nervous system is the system in the body that plays a leading role in regulating physiological functions and activities. It is composed of nervous tissue and is divided into two major parts: the central nervous system and the peripheral nervous system. The central nervous system includes the brain and spinal cord, and the peripheral nervous system includes cranial nerves and spinal nerves. Medical imaging of the brain, spinal cord and other nervous systems includes three-dimensional imaging, CT, MRI (T1, T2, DWI, SWI), MRS, MRA, MRV, cranial and cerebrovascular DSA and PET, etc. The medical imaging of the nervous system is the most complex part of medical imaging. Due to the various differences in neurological lesions, many of the early clinical manifestations of nervous system diseases are atypical, accompanied by the etiology, pathology and clinical symptoms of common nervous system diseases, which makes the learning and teaching content of the nervous system extensive and the subject foundation complex. It is a difficult point and key point in the study of medical imaging, and it is easy to misdiagnose during diagnosis, making it difficult for young and middle-aged doctors and grassroots medical staff to grasp.

[0003] The current uneven distribution of medical resources, information asymmetry, and individual resource limitations have resulted in a lack of effective means for medical staff and other personnel with relevant professional knowledge in the clinical diagnosis and learning practice of neurological medical imaging. In addition, the complexity of nervous system regulation, large individual differences, poor direct observability, and low collection rates of pathological specimens and biopsies have always been difficult points in medical imaging diagnosis and treatment. Artificial intelligence is a cutting-edge discipline that uses computer technology to simulate the human thinking and learning process. Currently, more than 90% of medical data comes from medical imaging. Due to the relatively simple data structure of medical images and the relatively mature image recognition technology, it has promoted the rapid development of artificial intelligence in the field of medical imaging diagnosis.

[0004] However, the training of artificial intelligence algorithm models requires a huge amount of manually labeled data, especially the above-mentioned neurological medical images, which face the challenges of scarce medical data and difficult labeling. Intelligence has not yet been achieved in data collection and data set production. In addition, the current conventional teaching software is only for mass teaching. There is no teaching software for professional medical imaging, especially complex neurological medical images. There is a lack of access to, operation of professional imaging data and professional examinations. Therefore, there is an urgent need to develop an efficient and accurate artificial intelligence-assisted neurological medical imaging teaching system for neurological medical imaging education, learning, and examinations, and to provide teaching and training for neurological medical imaging doctors. Summary of the invention

[0005] The purpose of the present invention is to provide a system for building a nervous system medical imaging case database based on artificial intelligence to solve the above problems.

[0006] To achieve the above object, the present invention provides the following technical solutions: A nervous system medical imaging case database construction system based on artificial intelligence, comprising a nervous system medical imaging database module, a teaching interactive learning platform module, a user self-test module and a user online examination module, wherein the nervous system medical imaging database module is signal-connected to the teaching interactive learning platform module, the teaching interactive learning platform module is signal-connected to the user self-test module, and the user self-test module is signal-connected to the user online examination module; The neurological medical imaging database module uses a 3D U-net neural network to perform lesion segmentation on the input neurological medical imaging data to obtain a target image, uses machine learning to extract image features of the target image, uses deep learning to extract medical image density, shape, and texture features, builds a neurological imaging prediction model, predicts test effect evaluation, creates a disease classification model and evaluates the classification results, corrects and improves the classification results of the evaluated image data through image expert interpretation feedback, re-enters the improved image data into the neurological medical imaging case database, and establishes a neurological medical imaging case database; The teaching interactive learning platform module supports users to learn medical imaging data and reinforce learning content in the nervous system medical imaging case database online, and intelligently and real-timely feedback the user's learning level and records and analyzes the user's test results, and pushes the same type of teaching content to the user multiple times according to the user's learning level; The user self-test module is used to test the user's learning results; The user online examination module is used to conduct online examinations on users.

[0007] Preferably, in the nervous system medical image database module, when a 3D U-net neural network is used to perform lesion segmentation on the input nervous system medical image data, the segmentation model adopts a U-net network, and the U-net network includes a feature extraction part and an upsampling part.

[0008] Preferably, machine learning is used to extract image features from the target image, and deep learning is used to extract density, shape, and texture features of medical images to construct a neuroimaging prediction model, including the following steps: A1. Data preparation: Preprocess the target image, including noise removal, normalization, image enhancement, and registration of images of different modalities; A2. Selection and construction of deep learning model: Convolutional neural network is selected as the deep learning model for extracting image features. Its basic components include convolution layer, pooling layer and fully connected layer. According to the task requirements and data characteristics, the appropriate CNN model architecture is selected, and the feature extraction layer is designed to extract image features; A3. Model training classification: Divide the preprocessed target image data into training set, validation set and test set, determine the training parameters of the model, and select the appropriate loss function according to the type of prediction task. For classification tasks, use the cross entropy loss function, and for regression tasks, select the mean square error loss function. Input the training set data into the constructed deep learning model, calculate the prediction result of the model through forward propagation, calculate the error between the prediction result and the true label according to the loss function, and then calculate the gradient through the back propagation algorithm to update the model parameters. During the training process, regularly use the validation set to evaluate the performance indicators of the model, and adjust the model's hyperparameters according to the evaluation results of the validation set; A4. Model evaluation: Use the test set to perform a final evaluation on the trained model, calculate the various performance indicators of the model on the test set, determine the actual predictive ability of the model, and evaluate the model performance by drawing the receiver operating characteristic curve and precision-recall curve.

[0009] A5. Model optimization: Based on the evaluation results, the model is further optimized. For the problem of model overfitting, regularization methods are used to constrain the complexity of the model. At the same time, the model architecture is adjusted. By comparing different model construction and feature extraction methods, the optimization plan is continuously fed back.

[0010] Preferably, in step A2, designing a feature extraction layer to extract image features includes the following steps: A21. Density feature extraction: In the convolution layer, the size and step size of the convolution kernel are adjusted to capture the information of different density areas in the image and normalize the features output by the convolution layer. A22. Shape feature extraction: Use convolution kernels of different shapes to perform convolution operations on the image, extract shape information from multiple angles, introduce dilated convolution, and expand the receptive field of the convolution kernel by adjusting the dilation rate; A23. Texture feature extraction: Multi-scale convolution operations are used to extract textures from small-scale detail textures to larger-scale macro textures. Gaussian kernels of different sizes are used for convolution to obtain texture feature maps of different scales. Combined with the idea of ​​gray-level co-occurrence matrix, operations that can capture the spatial relationship between pixels are designed in the convolution layer.

[0011] Preferably, the prediction test effect evaluation, creating a disease classification model and evaluating the classification results, includes the following steps; B1. Using the logistic regression method, we established a semi-parametric model of the expected outcome under the assumption of maximizing the consistency between the actual and predicted outcomes, constructed a loss function, minimized the estimated model comparison parameters, and realized the automatic classification of disease categories. B2. Evaluate the feature screening method and machine learning model in the test set and give the disease classification results.

[0012] Preferably, the classification results of the evaluated image data are corrected and improved through the interpretation feedback of image experts, and the improved image data are re-entered into the nervous system medical imaging case database, including the following steps; C1. Establish an information security mechanism based on data access and connection of hospital information systems to collect, screen and classify relevant imaging and clinical information of diseases; C2. Based on artificial intelligence, automatic segmentation, feature extraction and model training of lesions are carried out, and imaging information is integrated with clinical, pathological and follow-up information to establish a neuroimaging teaching database.

[0013] Preferably, the teaching interactive learning platform module includes an online learning module, a consolidation and reinforcement module, an intelligent interactive question-answering module, an intelligent feedback module, a testing module and a personalized learning path planning module; The online learning module supports users to learn online the medical imaging data in the nervous system medical imaging case database. After entering the online learning module, users can retrieve various types of nervous system medical imaging images, pre-process the images, annotate the input and identify structural lesions, detect and segment the image edges, measure and reconstruct the images, learn by viewing the medical imaging features of the disease, and finally obtain the results of disease diagnosis through label input; The consolidation and reinforcement module supports users to strengthen learning content. The consolidation and reinforcement module provides visualized three-dimensional medical images and intelligent engine libraries, automatically matches the best features with gold standard data, and conducts interactive learning. Users can describe, analyze, and predict images as needed, obtain corresponding features through neural networks, and intelligently push cases with similar attributes, so that users can assist in strengthening learning training; The intelligent interactive question-answering module supports users to ask questions in the learning process by text, voice or uploading images. The intelligent interactive question-answering module uses natural language processing technology to understand questions and gives answers by searching the nervous system medical imaging case database and artificial intelligence algorithms. At the same time, it automatically associates relevant knowledge points and learning resources according to the questions; During the user's learning process, the intelligent feedback module collects the user's learning answer data, image analysis exercise result data, and learning time data, tests and records the learning process, evaluates the user's mastery through a classification model, increases the weight ratio of the content that the user has not mastered, grasps the user's learning level, and provides real-time intelligent feedback, pushing the same type of teaching content to the user multiple times; The test module is used to record and analyze the test results of the user. The test module automatically outputs a test report, analyzes and evaluates the test results, records the test results of each user's multiple learning, analyzes and outputs the knowledge points that have not been mastered, and transmits them to the intelligent feedback module for subsequent positive feedback and negative feedback; The personalized learning path planning module tailors a personalized learning path for the user according to the user's learning level in the intelligent feedback module.

[0014] Preferably, when testing the user's learning results, the user self-test module randomly selects a certain number of cases, and the user performs online film reading and diagnosis. After submission, the user self-test module automatically scores and provides feedback, records the test time, and forms a self-test learning curve.

[0015] Preferably, when the user online examination module conducts an online examination on the user, the image pictures of the case can be presented on the examination paper, and it supports setting a variety of examination question types, including single choice, multiple choice, fill-in-the-blank, true-or-false questions, definition of terms and essay questions. When marking the papers, the user online examination module can set the system to automatically score or manually score, supports printing and exporting of examination results, and analyzes and summarizes the results.

[0016] Beneficial effects of the present invention: 1. In the present invention, the collection process of medical image data strictly guarantees the original information of the data. The collection of image examination data is directly oriented to the equipment, and there is no data circulation in the middle, which ensures that the original information of the data will not be tampered with. The image teaching gallery of this application is imported in batches, test papers are randomly / manually generated, and test papers are automatically graded and analyzed to generate archived test papers. The professional neurological medical imaging teaching provided by this application is mainly based on actual case teaching, and key images can be accessed. It is rich in content, taking into account theoretical knowledge, case image analysis, film reading and report writing exercises, and diagnostic thinking training. The test can be started in real time after the teaching is completed, and the evaluation of learning effects is more comprehensive and systematic.

[0017] 2. In the present invention, the system can be used for the teaching of complex neurological medical imaging, mobilize the enthusiasm and initiative of imaging medical students and resident training users, provide real-time feedback on learning effects through computers, promote the rapid integration of theoretical knowledge and clinical practice, facilitate teachers to prepare lessons accurately, and realize a flexible and convenient education model and a precise and individualized training method, which is conducive to accelerating the training process of imaging resident physicians, improving teaching efficiency and quality, and adapting to modern teaching methods.

[0018] 3. In the present invention, the teaching system truly restores the writing of diagnostic reports by doctors in the clinical working environment, organically combines book knowledge with clinical practice, and has a deeper knowledge learning. It ensures the natural connection and continuity of the training process of medical imaging students from school to hospital, and lays a solid foundation for the real entry into the writing of clinical diagnostic reports in the future, improves the teaching quality and practical skills, reduces the teaching burden of clinical teachers, effectively reduces the cost of teaching and training, and realizes the medical imaging education from digitalization and informatization to intelligence. The medical imaging teaching system of this application breaks the limitations of time and space, which can not only help develop local teaching in hospitals, but also can be promoted nationwide, realize cross-regional and cross-school resource sharing, and solve the problem of uneven educational resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The schematic diagram of the artificial intelligence-based nervous system medical imaging case database construction system of the present application is shown; Figure 2 The schematic diagram of establishing a nervous system medical imaging case database of the present application is shown; Figure 3 The schematic diagram of the teaching interactive learning platform module of the present application is shown; Figure 4 The schematic diagram of the online learning module in the teaching interactive learning platform module of the present application is shown; Figure 5 The schematic diagram of the consolidation and reinforcement module in the teaching interactive learning platform module of the present application is shown; Figure 6The schematic diagram of the intelligent feedback module in the teaching interactive learning platform module of the present application is shown; Figure 7 The schematic diagram of the test module in the teaching interactive learning platform module of the present application is shown; Figure 8 A diagram showing the internal structure of a computer device in the illustrated embodiment.

[0020] In the figure: 1. Neurological medical imaging database module; 2. Teaching interactive learning platform module; 3. User self-test module; 4. User online examination module; 21. Online learning module; 22. Consolidation and reinforcement module; 23. Intelligent interactive question-and-answer module; 24. Intelligent feedback module; 25. Testing module; 26. Personalized learning path planning module. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The following embodiments and experimental examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0022] This application provides a system for building a neurological medical imaging case database based on artificial intelligence, such as Figure 1 As shown, the system includes a nervous system medical image database module 1, a teaching interactive learning platform module 2, a user self-test module 3 and a user online examination module 4. The nervous system medical image database module 1 is signal-connected to the teaching interactive learning platform module 2, the teaching interactive learning platform module 2 is signal-connected to the user self-test module 3, and the user self-test module 3 is signal-connected to the user online examination module 4.

[0023] Figure 2 The schematic diagram of establishing a nervous system medical imaging case database of the present application is shown, and the present application is further explained below in conjunction with specific embodiments and experimental examples; When establishing a nervous system medical imaging case database, the nervous system medical imaging database module 1 first inputs nervous system medical imaging and collects medical imaging picture data of nervous system diseases stored in the actual clinical medical information system, including but not limited to three-dimensional imaging, CT, MRI (T1, T2, DWI, SWI), MRS, MRA, MRV, cranial cerebrovascular DSA and PET and other medical imaging pictures; Secondly, using 3D The U-net neural network performs lesion segmentation on the input nervous system medical image data to obtain a target image; wherein, the MRI brain nervous system image itself is a three-dimensional data structure, and the traditional 2D network-based segmentation algorithm needs to slice it, and the sliced ​​data is 2D data, which makes it impossible for the segmentation algorithm to learn the structural relationship between the layers of the data, resulting in insufficient network model learning, and the subsequent simple stacking of the 2D slice segmentation results may form jagged or fault phenomena, and the spatial feature expression of the lesion is missing. The present application uses a 3D fully convolutional neural network for image segmentation, and the segmentation model uses a U-net network. The amount of medical image data of the nervous system is relatively small, the U-net network model is small in size, and has fewer parameters, which is suitable for medical image segmentation, and the U-net network includes a feature extraction part and an upsampling part. For the feature extraction part, each pooling layer has a scale, including a total of 5 scales including the original image scale. For the upsampling part, each upsampling is fused with the same scale as the number of channels corresponding to the feature extraction part, and the low-resolution information during downsampling is combined to provide a basis for object category recognition and the high-resolution information during upsampling provides a basis for accurate segmentation and positioning, and fills in the underlying information to improve the image segmentation accuracy; Third, machine learning is used to extract image features from target images. Specifically, deep learning is used to extract density, shape, and texture features of medical images and to build a neural imaging prediction model. Among them, through automatic segmentation, feature extraction, model training and effect evaluation, the lesions are segmented to form training sets, validation sets and test sets; the density, volume, relative position, number and other parameters of the lesions are measured to generate and extract features; the extracted features are trained and classified into models; finally, different model construction and feature extraction methods are compared, and optimization solutions are continuously fed back to improve the effects of image feature prediction and disease classification; Fourth, predictive test effect evaluation, creating a disease classification model and evaluating the classification results; Among them, the logistic regression method is used to establish a semi-parametric model of the expected value of the outcome under the assumption of maximizing the consistency between the actual and predicted outcomes, construct a loss function, minimize the estimated model comparison parameters, and realize automatic classification of disease categories, such as bleeding, ischemia, tumors and other diseases; the feature screening method and machine learning model are evaluated in the test set to give the classification results of the disease; Finally, the classification results of the evaluated image data are corrected and improved through the interpretation feedback of imaging experts, and the improved image data are re-entered into the neurological medical imaging case database to establish the neurological medical imaging case database; In a preferred embodiment, by establishing an information security mechanism based on hospital information system data access and connection, relevant imaging and clinical information of diseases are collected, screened and classified, including 7 diseases required to be mastered in the standardized training outline for resident physicians, totaling 2,800 cases (including 800 cases of craniocerebral trauma; 700 cases of brain tumors; 500 cases of intraspinal tumors; 300 cases of cerebrovascular diseases; 300 cases of brain and spinal cord infectious diseases; and 200 cases of demyelinating diseases). Based on artificial intelligence, automatic segmentation, feature extraction, and model training of lesions are performed, and imaging information is integrated with clinical, pathological, and follow-up information to establish a high-quality neuroimaging teaching database.

[0024] Figure 3 The teaching interactive learning platform module 2 of the present application is shown in the figure, and the present application is further explained below in conjunction with specific embodiments and experimental examples; The teaching interactive learning platform module 2 supports users to learn online the medical imaging data and reinforced learning content in the nervous system medical imaging case database, and at the same time intelligently and real-timely feedbacks the user's learning level and records and analyzes the user's test results. According to the user's learning level, the same type of teaching content is pushed to the user multiple times. The teaching interactive learning platform module 2 includes an online learning module 21, a consolidation and reinforcement module 22, an intelligent interactive question-answering module 23, an intelligent feedback module 24, a test module 25 and a personalized learning path planning module 26; Figure 4 The online learning module 21 in the teaching interactive learning platform module 2 of the present application is shown, wherein in the online learning module 21, after entering the online learning module 21, the user retrieves various types of medical imaging images of the nervous system, pre-processes the images, annotates the input and identifies the structural lesions, detects and segments the image edges, measures and reconstructs the images, learns by viewing the medical imaging features of the disease, and finally obtains the results of disease diagnosis through label input, thereby helping the user to master imaging knowledge; Figure 5 The consolidation and reinforcement module 22 in the teaching interactive learning platform module 2 of the present application shown in the consolidation and reinforcement module 22 provides a visualized three-dimensional medical image and an intelligent engine library, automatically matches the best features of the gold standard data, and conducts interactive learning. The user describes, analyzes, and predicts the image according to the needs, obtains the corresponding features through the neural network, and intelligently pushes cases with similar attributes, which is convenient for users to assist in reinforcement learning training; Figure 6The intelligent feedback module 24 in the teaching interactive learning platform module 2 of the present application shown in the figure, during the user's learning process, collects the user's learning answer data, image analysis exercise result data and learning time data, tests and records the learning process with feedback, evaluates the user's mastery through the classification model, increases the weight ratio of the content that the user has not mastered, grasps the user's learning level, and provides real-time intelligent feedback, pushing the same type of teaching content to the user multiple times; Figure 7 The test module 25 in the teaching interactive learning platform module 2 of the present application is shown. In the test module 25, a test report is automatically output, the test results are analyzed and evaluated, the test results of each user's multiple learning are recorded, the knowledge points that have not been mastered are analyzed and output, and transmitted to the intelligent feedback module 24 for subsequent positive feedback and negative feedback.

[0025] When testing the user's learning results, the user self-test module 3 randomly selects a certain number of cases, and the user performs online film reading and diagnosis. After submission, the user self-test module 3 automatically scores and gives feedback, records the test time, and forms a self-test learning curve. The user can view the correct diagnosis results and case knowledge points for retrospective learning; through repeated film reading training on a large number of typical cases, the user can quickly complete the transition from text knowledge learned from textbooks to understanding image information and master film reading skills. At the same time, this module can evaluate the learning effect and progress of the user, analyze a large amount of historical data based on the user's performance, provide guidance for the user, customize personalized training, and ultimately achieve personalized education, greatly promoting the precision education of imaging learning; When the user's online examination module 4 takes an online examination for the user, the image pictures of the case can be presented on the test paper accurately and without distortion, which is close to the clinical practice and examination requirements of the imaging profession, making the medical imaging examination standardized, and supports the setting of various examination question types, including single-choice, multiple-choice, fill-in-the-blank, true-or-false questions, definition of terms and essay questions. When marking the paper, the user's online examination module 4 can set the system to automatically score or manually score. For objective questions, the system automatically scores, saving teachers' time and improving examination efficiency. It supports printing and exporting test scores, and analyzes and summarizes the scores.

[0026] In the present invention, the collection process of medical image data strictly guarantees the original information of the data. The collection of image examination data is directly oriented to the equipment, and there is no data circulation in the middle, which ensures that the original information of the data will not be tampered with. The image teaching library of this application is imported in batches, and test papers are randomly / manually generated, and the test papers are automatically graded and analyzed to generate archived test papers. The professional neurological medical imaging teaching provided by this application is mainly based on actual case teaching, and key images can be accessed. It is rich in content, taking into account theoretical knowledge, case image analysis, film reading and report writing exercises, and diagnostic thinking training. The test can be started in real time after the teaching is completed, and the evaluation of learning effect is more comprehensive and systematic. At the same time, the artificial intelligence-based neurological medical imaging case database construction system of this application can mobilize the enthusiasm and initiative of imaging medical students and resident training users for complex neurological medical imaging teaching, and promote the rapid integration of theoretical knowledge and clinical practice through real-time feedback of learning effects by computers, so as to facilitate teachers to prepare lessons accurately, realize a flexible and convenient education model and a precise and individualized training method, which is conducive to accelerating the training process of imaging resident physicians, improving teaching efficiency and quality, and adapting to modern teaching methods; The teaching system of this application truly restores the writing of diagnostic reports by doctors in the clinical working environment, organically combines book knowledge with clinical practice, and enables deeper knowledge learning. It ensures the natural connection and continuity of the training process of medical imaging students from school to hospital, and lays a solid foundation for truly entering the writing of clinical diagnostic reports in the future. It improves the teaching quality and practical skills, reduces the teaching burden of clinical teachers, effectively reduces the cost of teaching and training, and realizes the transformation of medical imaging education from digitalization and informatization to intelligence. The medical imaging teaching system of this application breaks the limitations of time and space, which can not only help develop local teaching in hospitals, but also be promoted nationwide to realize cross-regional and cross-school resource sharing and solve the problem of uneven educational resources.

[0027] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, a memory, a network interface and a database connected through a system bus, wherein the processor of the computer device is used to provide computing and control capabilities, the memory of the computer device includes a non-volatile storage medium and an internal memory, the non-volatile storage medium stores an operating system, a computer program and a database, the internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium, the database of the computer device is used to store configuration templates, and can also be used to store target web page data, the network interface of the computer device is used to communicate with an external terminal through a network connection, and the computer program is executed by the processor to realize a method for predicting the expansion and contraction deformation of a pipe joint of an immersed tube tunnel.

[0028] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0029] In one embodiment, a storage medium is also provided, wherein the storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a computer, the computer executes the system as described in the aforementioned embodiment, and the computer may be a part of the aforementioned system.

[0030] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be implemented using a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented using program codes executable by the computing device, so that they can be stored in a computer storage medium (ROM / RAM, magnetic disk, optical disk) and executed by the computing device. In some cases, the steps shown or described can be executed in a different order from that shown here, or they can be made into individual integrated circuit modules separately, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Therefore, the present application is not limited to any specific combination of hardware and software.

[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A system for building a nervous system medical imaging case database based on artificial intelligence, characterized by: It comprises a nervous system medical image database module (1), a teaching interactive learning platform module (2), a user self-test module (3) and a user online examination module (4), wherein the nervous system medical image database module (1) is signal-connected to the teaching interactive learning platform module (2), the teaching interactive learning platform module (2) is signal-connected to the user self-test module (3), and the user self-test module (3) is signal-connected to the user online examination module (4); The neurological medical imaging database module (1) uses a 3D U-net neural network to perform lesion segmentation on the input neurological medical imaging data to obtain a target image, uses machine learning to extract image features of the target image, uses deep learning to extract medical image density, shape, and texture features, builds a neurological imaging prediction model, predicts test effect evaluation, creates a disease classification model and evaluates the classification results, corrects and improves the classification results of the evaluated image data through image expert interpretation feedback, re-enters the improved image data into the neurological medical imaging case database, and establishes a neurological medical imaging case database; The teaching interactive learning platform module (2) supports users to learn online medical imaging data and reinforced learning content in the nervous system medical imaging case database, and at the same time intelligently and real-timely feedback the user's learning level and records and analyzes the user's test results, and pushes the same type of teaching content to the user multiple times according to the user's learning level; The user self-test module (3) is used to test the user's learning results; The user online examination module (4) is used to conduct an online examination on the user.

2. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: In the nervous system medical image database module (1), when a 3D U-net neural network is used to perform lesion segmentation on the input nervous system medical image data, the segmentation model uses a U-net network, and the U-net network includes a feature extraction part and an upsampling part.

3. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: Machine learning is used to extract image features from target images, and deep learning is used to extract density, shape, and texture features of medical images to build a neural imaging prediction model, including the following steps: A1. Data preparation: Preprocess the target image, including noise removal, normalization, image enhancement, and registration of images of different modalities; A2. Selection and construction of deep learning model: Convolutional neural network is selected as the deep learning model for extracting image features. Its basic components include convolution layer, pooling layer and fully connected layer. According to the task requirements and data characteristics, the appropriate CNN model architecture is selected, and the feature extraction layer is designed to extract image features; A3. Model training classification: Divide the preprocessed target image data into training set, validation set and test set, determine the training parameters of the model, and select the appropriate loss function according to the type of prediction task. For classification tasks, use the cross entropy loss function, and for regression tasks, select the mean square error loss function. Input the training set data into the constructed deep learning model, calculate the prediction result of the model through forward propagation, calculate the error between the prediction result and the true label according to the loss function, and then calculate the gradient through the back propagation algorithm to update the model parameters. During the training process, regularly use the validation set to evaluate the performance indicators of the model, and adjust the model's hyperparameters according to the evaluation results of the validation set; A4. Model evaluation: Use the test set to perform a final evaluation on the trained model, calculate the various performance indicators of the model on the test set, determine the actual prediction ability of the model, and evaluate the model performance by drawing the receiver operating characteristic curve and precision-recall curve; A5. Model optimization: Based on the evaluation results, the model is further optimized. For the problem of model overfitting, regularization methods are used to constrain the complexity of the model. At the same time, the model architecture is adjusted. By comparing different model construction and feature extraction methods, the optimization plan is continuously fed back.

4. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 3 is characterized in that: In step A2, a feature extraction layer is designed to extract image features, including the following steps: A21. Density feature extraction: In the convolution layer, the size and step size of the convolution kernel are adjusted to capture the information of different density areas in the image and normalize the features output by the convolution layer. A22. Shape feature extraction: Use convolution kernels of different shapes to perform convolution operations on the image, extract shape information from multiple angles, introduce dilated convolution, and expand the receptive field of the convolution kernel by adjusting the dilation rate; A23. Texture feature extraction: Multi-scale convolution operations are used to extract textures from small-scale detail textures to larger-scale macro textures. Gaussian kernels of different sizes are used for convolution to obtain texture feature maps of different scales. Combined with the idea of ​​gray-level co-occurrence matrix, operations that can capture the spatial relationship between pixels are designed in the convolution layer.

5. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: Predictive test effect evaluation, create disease classification models and evaluate classification results, The steps include: B1. Using the logistic regression method, we established a semi-parametric model of the expected outcome under the assumption of maximizing the consistency between the actual and predicted outcomes, constructed a loss function, minimized the estimated model comparison parameters, and realized the automatic classification of disease categories. B2. Evaluate the feature screening method and machine learning model in the test set and give the disease classification results.

6. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: The classification results of the evaluated image data are corrected and improved through the interpretation feedback of image experts, and the improved image data are re-entered into the nervous system medical imaging case database, including the following steps; C1. Establish an information security mechanism based on data access and connection of hospital information systems to collect, screen and classify relevant imaging and clinical information of diseases; C2. Based on artificial intelligence, automatic segmentation, feature extraction and model training of lesions are carried out, and imaging information is integrated with clinical, pathological and follow-up information to establish a neuroimaging teaching database.

7. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: The teaching interactive learning platform module (2) comprises an online learning module (21), a consolidation and reinforcement module (22), an intelligent interactive question-answering module (23), an intelligent feedback module (24), a testing module (25) and a personalized learning path planning module (26); The online learning module (21) supports users to learn online medical imaging data in a nervous system medical imaging case database. After entering the online learning module (21), the user retrieves various types of nervous system medical imaging images, pre-processes the images, labels the input and identifies structural lesions, detects and segments the image edges, measures and reconstructs the images, learns by viewing the medical imaging features of the disease, and finally obtains the result of disease diagnosis through label input; The consolidation and reinforcement module (22) supports users to strengthen learning content. The consolidation and reinforcement module (22) provides a visualized three-dimensional medical image and an intelligent engine library, automatically matches the best features of the gold standard data, and conducts interactive learning. The user describes, analyzes, and predicts the image according to needs, obtains corresponding features through a neural network, and intelligently pushes cases with similar attributes, so that the user can assist in strengthening learning training; The intelligent interactive question-answering module (23) supports users to ask questions in the learning process by text, voice or uploading images. The intelligent interactive question-answering module (23) uses natural language processing technology to understand the questions, and provides answers by searching the nervous system medical imaging case database and artificial intelligence algorithms, and automatically associates relevant knowledge points and learning resources according to the questions; The intelligent feedback module (24) collects the user's learning answer data, image analysis exercise result data and learning time data during the user's learning process, tests and records the learning process with feedback, evaluates the user's mastery through a classification model, increases the weight ratio of the content that the user has not mastered, grasps the user's learning level, and provides real-time intelligent feedback, pushing the same type of teaching content to the user multiple times; The test module (25) is used to record and analyze the test results of the user. The test module (25) automatically outputs a test report, analyzes and evaluates the test results, records the test results of each user's multiple learning, analyzes and outputs the knowledge points that have not been mastered, and transmits them to the intelligent feedback module (24) for subsequent positive feedback and negative feedback; The personalized learning path planning module (26) tailors a personalized learning path for the user according to the learning level of the user in the intelligent feedback module (24).

8. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: When testing the user's learning results, the user self-test module (3) randomly selects a certain number of cases, and the user performs online film reading and diagnosis. After submission, the user self-test module (3) automatically scores and provides feedback, records the test time, and forms a self-test learning curve.

9. The artificial intelligence-based nervous system medical imaging case database construction system according to claim 1 is characterized in that: When the user online examination module (4) conducts an online examination on the user, images of the case can be presented on the examination paper, and a variety of examination question types can be set, including single-choice, multiple-choice, fill-in-the-blank, true-or-false, definition of terms and essay questions. When marking the paper, the user online examination module (4) can set the system to automatically score or manually score, and supports printing and exporting of examination results, and analyzing and summarizing the results.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the system according to any one of claims 1 to 9 are implemented.

Images