[go: up one dir, main page]

WO2021031279A1 - 一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法 - Google Patents

一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法 Download PDF

Info

Publication number
WO2021031279A1
WO2021031279A1 PCT/CN2019/107225 CN2019107225W WO2021031279A1 WO 2021031279 A1 WO2021031279 A1 WO 2021031279A1 CN 2019107225 W CN2019107225 W CN 2019107225W WO 2021031279 A1 WO2021031279 A1 WO 2021031279A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
pneumonia
layer
picture
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/107225
Other languages
English (en)
French (fr)
Inventor
田宜聪
任涛
王逸群
刘昕靓
杨丹丹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeastern University China
Original Assignee
Northeastern University China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northeastern University China filed Critical Northeastern University China
Publication of WO2021031279A1 publication Critical patent/WO2021031279A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/25—Integrating or interfacing systems involving database management systems
    • G06F16/252—Integrating or interfacing systems involving database management systems between a Database Management System and a front-end application
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28—Databases characterised by their database models, e.g. relational or object models
    • G06F16/284—Relational databases
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Definitions

  • the invention relates to the technical field of medical image classification, in particular to an X-Ray chest radiograph pneumonia intelligent diagnosis system and method based on deep learning.
  • Imaging is one of the important information in disease diagnosis. Imaging technologies include: Computed Tomography (CT), Positron Emission Tomography (PET), Single Photon Radiation Tomography, and MRI Imaging, ultrasound imaging, etc.
  • CT Computed Tomography
  • PET Positron Emission Tomography
  • MRI Imaging Magnetic Imaging Imaging
  • ultrasound imaging etc.
  • CT Computed Tomography
  • PET Positron Emission Tomography
  • MRI Imaging Magnetic Imaging Imaging
  • ultrasound imaging etc.
  • traditional X-ray diagnosis is still one of the important basis for radiological diagnosis.
  • the chest cavity is hailed as the mirror of human health and disease, because it contains many important tissue structures of the human body, and can provide various information about the human body, such as the diagnosis of lung diseases, rib fractures and injuries, heart enlargement symptoms, cardiopulmonary coefficients, etc. All can be identified and confirmed by X-Ray chest film. Chest X-ray is still a routine examination item for hospital imaging diagnosis.
  • X-Ray chest radiographs account for more than 40% of all radiographic diagnosis due to its low price and weak radiation dose. This reflects the important application value of X-Ray chest film in the medical field.
  • Pranav et al. proposed a deep learning-based network CheXNet that can assist in the diagnosis of lung X-Ray images.
  • Guan et al. used deep learning to integrate the attention model to assist in diagnosis of lung X-Ray images.
  • Most of the above are related studies on lung nodules. Even if there are related studies on the automatic recognition of pneumonia, it is based on the deep learning of the lung X-Ray multi-classification.
  • the intelligent diagnosis of pneumonia lacks pertinence.
  • the focus area of pneumonia is easy Obscured by the normal tissue structure, it is difficult to distinguish the fine pneumonia lesions from the tissue structure, and it is easy to miss the pneumonia lesions in the detection results of the lesions.
  • the use of traditional computer-aided diagnosis methods often cannot achieve high accuracy.
  • the present invention provides an X-Ray chest radiograph pneumonia intelligent diagnosis system and method based on deep learning.
  • an X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning includes three modules: user login and registration module, image format conversion module and Pneumonia prediction module, the brief business process of the system is shown in Figure 2;
  • the user login and registration module provides login function, registration function and reset password function, which is used to provide users with the entrance to the system, and the process is shown in Figure 3;
  • the login function requires the user to enter the account number and password on the login interface and submit them to the system.
  • the back end of the system queries the user information table of the database for the corresponding password according to the entered account number. If the returned result is empty, the account number entered by the user is indicated. Does not exist. If the returned result does not match the password entered by the user, it means that the user password is entered incorrectly. Only when the password entered by the user matches the password queried from the database, the system will display the corresponding jump interface;
  • the registration function is that the user can enter the account number, password, phone number, and email address to register.
  • the background will judge the legality of this information through a JS script. If the information is all legal, a new piece of data is added to the user information table of the database;
  • the password reset function is that when the user forgets the password, he can enter the account and email verification information on the password recovery page. When the email verification information is correct, the system will allow the user to reset the password and modify the corresponding information in the user information table. Password information.
  • the picture format conversion module provides the function of selecting local pictures and format conversion.
  • the format conversion function is used to convert the dicom format data of the image center into the jpeg format required by the system. The process is shown in Figure 4.
  • the pneumonia prediction module provides patient information addition and picture prediction functions, which are used to enter patient information and add corresponding image data to the patient database, and then use the trained prediction model in the system to predict the patient's data Analysis, the process is shown in Figure 5;
  • the patient information is added by the user by entering the patient's name, gender, age, and date of visit into the system, and finally adding the picture to the patient information, the background will perform JS verification on these data, and then save it in the database.
  • the picture prediction function is to select a local picture and predict whether the patient suffers from pneumonia through the prediction model trained in the system.
  • the system uses the B/S architecture to divide the system into four layers, which are platform layer, support layer, service layer, and application layer.
  • the architecture is shown in Figure 1;
  • the application layer includes a system call interface, a Web access interface and a result visualization interface, which are connected to the user side;
  • the service layer includes user operation interfaces for user registration, user authentication, patient import, picture import, user login, format conversion, model loading, and picture prediction; among them, user registration, user login, and user authentication belong to the services provided by the login and registration module ; Format conversion belongs to the service provided by the picture format conversion module; patient import, picture import, model loading, and picture prediction belong to the service provided by the pneumonia prediction module;
  • the support layer includes classification methods based on deep convolutional neural networks, database management systems based on relational databases, traditional image processing methods, and medical image processing methods; among them, the classification methods based on deep convolutional neural networks and traditional image processing methods are image prediction Provide services; database management systems based on relational databases provide services for user registration, user login, and user authentication; medical image processing methods provide services for format conversion;
  • the platform layer adopts the Keras framework, and designs the deep convolutional neural network according to the convolutional layer, the BN layer, the advanced activation function layer, the pooling layer, the fully connected layer, and the softmax layer.
  • the loss function binary cross-entropy and the optimization function RMSProp are selected.
  • Convolutional neural network is optimized; sqlite3 relational database is used as the database management system of this system; ITK software library is used as the platform of medical image processing methods; the opencv computer vision library is used as the platform of traditional image processing methods.
  • the technical solution adopted by the present invention is: a method for diagnosis using an X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning.
  • the process of the method is shown in Figure 6 and includes the following steps:
  • Step 1 Obtain an X-Ray chest radiograph image in dicom format, call the ReadImage method in the ITK library to read the dicom format picture, then call the GetArrayFromImage method in ITK to extract the pixel matrix in the dicom image, and finally pass the imwrite in opencv The method saves the extracted pixel matrix into a picture in jpeg format;
  • Step 2 Select the dataset Chest X-Ray Images (Pneumonia), generate training set and test set;
  • Step 3 Establish a deep convolutional neural network VGG prediction model, as shown in Figure 7.
  • the VGG model includes six convolutional layers, BN layer, advanced activation function, two fully connected layers and the final softmax layer, and set the model The value of the number of training iterations epoch;
  • the weights trained by ImageNet are used to transfer the model.
  • a BatchNormalization layer is added between the convolutional layer and the activation function to speed up the network convergence.
  • the advanced activation function LeakyRelu replaces Relu, and the cross-entropy loss function binary cross- Entropy is used as the model optimization index, and the optimization function adopts RMSProp method to accelerate model convergence;
  • Step 4 Use the training set to train the VGG prediction model
  • Step 5 Use data enhancement methods such as flipping, rotation, and affine transformation to enhance the image data of the training set to obtain a new training set;
  • Step 6 Input the test set and test the trained VGG prediction model to obtain the prediction accuracy
  • Step 7 Repeat steps 4 to 7 to perform iterative training on the VGG prediction model of the deep convolutional neural network, until the number of iterations reaches the preset epoch value, stop the iteration;
  • Step 8 Save the VGG model with the highest accuracy on the test set
  • Step 9 Input the picture in jpeg format obtained in Step 1 into the VGG model with the highest accuracy saved in Step 8, to obtain the classification prediction result of the picture.
  • the present invention accelerates the convergence speed of model training through data enhancement, migration learning, and improvement of network structure, and improves the accuracy of pneumonia recognition and the generalization ability of the model;
  • the present invention also provides a Django-based BS application, which simplifies the user installation process and reduces the dependence on the local environment. Through friendly interface and system modules designed around actual needs, this system can be easily accepted by various hospitals, and it can also be easily integrated with other hospital systems.
  • the application based on MVC architecture also improves the efficiency of later maintenance and redevelopment;
  • Figure 1 is an architecture diagram of an X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning of the present invention
  • Figure 2 is a brief business flow chart of an X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning of the present invention
  • Figure 3 is a flow chart of the user login and registration module of the present invention.
  • FIG. 4 is a flowchart of the picture format conversion module of the present invention.
  • Figure 5 is a flowchart of the pneumonia prediction module of the present invention.
  • Fig. 6 is a flow chart of an X-Ray chest radiograph pneumonia intelligent diagnosis method based on deep learning of the present invention
  • FIG. 7 is a structural diagram of the VGG prediction model of the deep convolutional neural network of the present invention.
  • Figure 8 is a diagram of a user login interface in an embodiment of the present invention.
  • Figure 9 is a user registration interface diagram in an embodiment of the present invention.
  • Figure 10 is a picture format conversion interface diagram in an embodiment of the present invention.
  • Figure 11 is an interface diagram of adding patient information and uploading patient X-Ray chest radiographs and performing prediction in an embodiment of the present invention
  • Fig. 12 is an output interface for predicting pneumonia of a patient's X-Ray chest radiograph in an embodiment of the present invention.
  • An X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning The system includes three modules: user login and registration module, picture format conversion module, and pneumonia prediction module. The brief business process of the system is shown in Figure 2;
  • the user login and registration module provides login function, registration function and reset password function, which is used to provide users with the entrance to the system, and the process is shown in Figure 3;
  • the login function requires the user to enter the account number and password in the login interface and submit them to the system.
  • the user login interface is shown in Figure 8.
  • the system backend queries the user information table of the database for the corresponding password according to the entered account number. If the result is returned If it is empty, it means that the account entered by the user does not exist. If the result of returning the goods does not match the password entered by the user, it means that the user password is entered incorrectly. Only when the password entered by the user matches the password queried from the database, the system The corresponding jump interface will be displayed;
  • the registration function is that the user can enter the account number, password, phone number, and email address to register.
  • the user registration interface is shown in Figure 9.
  • the background will use the JS script to determine the legality of this information. If the information is all legal, the user in the database A new piece of data is added to the information table;
  • the password reset function is that when the user forgets the password, he can enter the account and email verification information on the password recovery page. When the email verification information is correct, the system will allow the user to reset the password and modify the corresponding information in the user information table. Password information.
  • the picture format conversion module provides functions for selecting local pictures and format conversion.
  • the picture format conversion interface is shown in Figure 10.
  • the format conversion function is used to convert the dicom format data of the image center into the jpeg format required by the system.
  • the pneumonia prediction module provides patient information addition and picture prediction functions, which are used to enter patient information and add corresponding image data to the patient database, add patient information and upload patient X-Ray chest radiographs and make predictions.
  • the interface is as shown in the figure 11, and then use the predicted model trained in the system to predict and analyze the patient’s data;
  • the patient information is added by the user by entering the patient's name, gender, age, and date of visit into the system, and finally adding the picture to the patient information, the background will perform JS verification on these data, and then save it in the database.
  • the picture prediction function is to select a local picture and predict whether the patient suffers from pneumonia through the prediction model trained in the system.
  • the system uses the B/S architecture to divide the system into four layers, which are platform layer, support layer, service layer, and application layer.
  • the architecture is shown in Figure 1;
  • the application layer includes a system call interface, a Web access interface and a result visualization interface, which are connected to the user side;
  • the service layer includes user operation interfaces for user registration, user authentication, patient import, picture import, user login, format conversion, model loading, and picture prediction; among them, user registration, user login, and user authentication belong to the services provided by the login and registration module ; Format conversion belongs to the service provided by the picture format conversion module; patient import, picture import, model loading, and picture prediction belong to the service provided by the pneumonia prediction module;
  • the support layer includes classification methods based on deep convolutional neural networks, database management systems based on relational databases, traditional image processing methods, and medical image processing methods; among them, the classification methods based on deep convolutional neural networks and traditional image processing methods are image prediction Provide services; database management systems based on relational databases provide services for user registration, user login, and user authentication; medical image processing methods provide services for format conversion;
  • the platform layer adopts the Keras framework, and designs the deep convolutional neural network according to the convolutional layer, the BN layer, the advanced activation function layer, the pooling layer, the fully connected layer, and the softmax layer.
  • the loss function binary cross-entropy and the optimization function RMSProp are selected.
  • Convolutional neural network is optimized; sqlite3 relational database is used as the database management system of this system; ITK software library is used as the platform of medical image processing methods; the opencv computer vision library is used as the platform of traditional image processing methods.
  • the technical solution adopted by the present invention is: a method for diagnosis using an X-Ray chest radiograph pneumonia intelligent diagnosis system based on deep learning.
  • the process of the method is shown in Figure 6 and includes the following steps:
  • Step 1 Obtain an X-Ray chest radiograph image in dicom format, call the ReadImage method in the ITK library to read the dicom format picture, then call the GetArrayFromImage method in ITK to extract the pixel matrix in the dicom image, and finally pass the imwrite in opencv The method saves the extracted pixel matrix into a picture in jpeg format;
  • Step 2 Select the dataset Chest X-Ray Images (Pneumonia), generate training set and test set;
  • the weights trained by ImageNet are used to transfer the model.
  • a BatchNormalization layer is added between the convolutional layer and the activation function to speed up the network convergence.
  • the advanced activation function LeakyRelu replaces Relu, and the cross-entropy loss function binary cross- Entropy is used as the model optimization index, and the optimization function adopts RMSProp method to accelerate model convergence;
  • Step 4 Use the training set to train the VGG prediction model
  • Step 5 Use data enhancement methods such as flipping, rotation, and affine transformation to enhance the image data of the training set to obtain a new training set;
  • Step 6 Input the test set and test the trained VGG prediction model to obtain the prediction accuracy
  • Step 7 Repeat steps 4 to 7 to perform iterative training on the VGG prediction model of the deep convolutional neural network, until the number of iterations reaches the preset epoch value, stop the iteration;
  • Step 8 Save the VGG model with the highest accuracy on the test set
  • Step 9 Input the picture in jpeg format obtained in Step 1 into the VGG model with the highest accuracy saved in Step 8, to obtain the classification prediction result of the picture.
  • the prediction result output rule is set: 1 is pneumonia, and 0 is normal;
  • the output of the prediction result of the X-Ray chest radiograph image of the patient is 0, and it can be judged that the patient is not suffering from pneumonia.
  • the result output interface is shown in FIG. 12.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Security & Cryptography (AREA)
  • Public Health (AREA)
  • Evolutionary Computation (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Pathology (AREA)
  • Computer Hardware Design (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法，属于医学图像分类技术领域，该系统包括用户登录与注册模块、图片格式转换模块和肺炎预测模块，运用B/S架构，将系统分成四层，分别为平台层、支撑层、服务层和应用层。用户安装过程简单，减轻对本地环境的依赖。通过友好的界面以及围绕实际需求而设计的系统模块，可以使得本系统和其他系统进行功能上的融合。基于深度学习的X-Ray胸片肺炎智能诊断方法通过数据增强迁移学习以及对网络结构进行改进等手段加快了模型训练的收敛速度，提高了肺炎识别的准确率以及模型的泛化能力。可以大大减轻了人工阅片给医生带来的困扰。

Description

一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法 技术领域
本发明涉及医学图像分类技术领域，尤其涉及一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法。
背景技术
医学影像是疾病诊断中重要的信息之一，其中成像技术包括：计算机断层成像技术(Computed Tomography，CT)、正电子放射层析成像(Positron Emission Tomography，PET)、单光子辐射断层成像、核磁共振成像、超声成像等，传统的X射线诊断仍然是重要的放射诊断依据之一。胸腔被喻为人体健康与疾病的镜子，因为它包含了人体许多种重要的组织结构，可以提供人体多方面的信息，如肺部疾病的诊断、肋骨骨折及损伤、心脏扩大症状、心肺系数等都可由X-Ray胸片来识别与确认。胸部X线仍为医院影像诊断的常规检查项目，即使需要CT、MRI检查的病例，也常需要参考胸部X线检查的表现。X-Ray胸片以其低廉的价格和微弱的放射剂量占所有放射影像诊断的40％以上。这体现了X-Ray胸片的在医学领域的重要应用价值。
但是如何分析胸片是一项极具挑战性的任务，即使是经验丰富的专家也经常会感到棘手。随着近年来深度学习技术在计算机视觉以及图像分类、分割、识别等领域的巨大发展，科研人员在X-Ray胸片的计算机诊断方面提出了很多辅助诊断方法，Aramato等提出了一种基于多灰度值阔值的方法来提取感兴趣区域，然后利用感兴趣区域的九个特征进行简单的分类来提取结节。Awai等人利用形态学方法定位肺结节。Li等人延续sato的研究成果，并且设计了能够增强特定退织在影像上的表现而减弱剩下组织展现的选择性增强函数，这样使得感兴趣区域更加明显。Pranav等人提出了一种基于深度学习的网络CheXNet能够对肺部X-Ray图像进行辅助诊断。Guan等人通过深度学习融合注意力模型的方式来对肺部X-Ray图像进行辅助诊断。上述大多是针对肺部结节的相关研究，即使有对于肺炎 自动识别的相关研究也是基于深度学习的肺部X-Ray多分类，对肺炎的智能诊断缺少针对性，此外由于肺炎病灶区域很容易被正常的组织结构遮挡，使得细微的肺炎病灶和组织结构难以区分，病灶检测结果也很容易遗漏肺炎病灶。采用传统的计算机辅助诊断手段往往不能达到较高的准确率。
发明概述
技术问题
问题的解决方案
技术解决方案
针对上述现有技术的不足，本发明提供一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法。
为解决上述技术问题，本发明所采取的技术方案是：一种基于深度学习的X-Ray胸片肺炎智能诊断系统，系统包括三个模块分别为：用户登录与注册模块、图片格式转换模块和肺炎预测模块，系统的简要业务流程如图2所示；
其中，用户登录与注册模块提供登录功能、注册功能和重置密码功能，用于对用户提供进入系统的入口，其流程如图3所示；
所述登录功能需要用户在登录界面输入账号和密码并提交给系统，系统后端根据输入的账号，向数据库的用户信息表中查询相应的密码，如果返回结果为空，则说明用户输入的账号不存在，如果返回的结果与用户输入的密码不匹配，则说明用户密码输入错误，只有当用户输入的密码和从数据库中查询的密码相匹配时，系统才会显示相应的跳转界面；
所述注册功能是用户可以输入账号、密码、电话、邮箱地址进行注册，后台会通过JS脚本对这些信息进行合法性判断，如果信息全都合法则在数据库的用户信息表中新增一条数据；
所述重置密码功能是当用户忘记密码时，可以在找回密码页面输入账号和邮箱的验证信息，当邮箱验证信息正确时，系统会允许用户进行密码重置，并修改用户信息表中相应的密码信息。
其中，图片格式转换模块提供选择本地图片以及格式转换功能，格式转换功能用于将影像中心的dicom格式数据转换为本系统所需jpeg格式，其流程如图4所示 。
其中，肺炎预测模块提供病人信息添加以及图片预测功能，用于将病人信息进行录入并将对应的影像数据添加至病人数据库中，然后使用本系统中训练好的预测模型对该病人的数据进行预测分析，其流程如图5所示；
所述病人信息添加是用户通过将患者姓名、性别、年龄、就诊日期录入系统，最后将图片添加至病人信息中，后台会对这些数据进行JS校验，之后保存至数据库中。
所述图片预测功能是选择一张本地图片，通过系统内训练好的预测模型来完成对患者是否罹患肺炎的预测。
该系统运用B/S架构，将系统分成四层，分别为平台层、支撑层、服务层和应用层，其架构如图1所示；
其中，应用层包含系统调用接口、Web访问接口和结果可视化接口，与用户端相连接；
服务层包含用户注册、用户认证、病人导入、图片导入、用户登录、格式转换、模型加载和图片预测的用户可操作界面；其中，用户注册、用户登录、用户认证属于登录与注册模块提供的服务；格式转换属于图片格式转换模块提供的服务；病人导入、图片导入、模型加载、图片预测属于肺炎预测模块提供的服务；
支撑层包含基于深度卷积神经网络的分类方法、基于关系型数据库的数据库管理系统、传统图像处理方法、医学图像处理方法；其中基于深度卷积神经网络的分类方法、传统图像处理方法为图片预测提供服务；基于关系型数据库的数据库管理系统为用户注册、用户登录、用户认证提供服务；医学图像处理方法为格式转换提供服务；
平台层采用Keras框架，按照卷积层、BN层、高级激活函数层、池化层、全连接层以及softmax层对深度卷积神经网络进行设计，选用损失函数binary cross-entropy及优化函数RMSProp对卷积神经网络进行优化；采用sqlite3关系型数据库作为本系统的数据库管理系统；采用ITK软件库作为医学图像处理方法的平台；采用opencv计算机视觉库作为传统图像处理方法的平台。
为解决上述技术问题，本发明所采取的技术方案是：一种采用基于深度学习的X-Ray胸片肺炎智能诊断系统进行诊断的方法，该方法的流程如图6所示，包括如下步骤：
步骤1：获取一张dicom格式的X-Ray胸片图像，调用ITK库中的ReadImage方法读取dicom格式图片，之后通过调用ITK中GetArrayFromImage方法提取dicom图像中的像素矩阵，最后通过opencv中的imwrite方法将已经提取的像素矩阵保存成jpeg格式的图片；
步骤2：选择数据集Chest X-Ray Images(Pneumonia)，生成训练集和测试集；
步骤3：建立深度卷积神经网络VGG预测模型，如图7所示，其中VGG模型包括六个卷积层、BN层、高级激活函数，两个全连接层以及最后的softmax层，并设置模型训练迭代次数epoch的值；
对于模型的权重加载采用ImageNet训练好的权重对模型进行迁移学习，在卷积层和激活函数之间加入BatchNormalization层来加快网络收敛速度，高级激活函数LeakyRelu代替Relu，采用交叉熵损失函数binary cross-entropy来作为模型优化指标，优化函数采用RMSProp方法加快模型收敛；
步骤4：用训练集对VGG预测模型进行训练；
步骤5：采用翻转、旋转、仿射变换这些数据增强的方法，对训练集的图像数据进行增强处理，得到新的训练集；
步骤6：输入测试集，对训练好的VGG预测模型进行测试，得到预测准确率；
步骤7：重复执行步骤4至步骤7，对深度卷积神经网络VGG预测模型进行迭代训练，直到迭代次数达到预设的epoch的值，停止迭代；
步骤8：将测试集上准确率最高的VGG模型进行保存；
步骤9：将步骤1得到的jpeg格式的图片输入到步骤8保存的准确率最高的VGG模型中，得到图片的分类预测结果。
发明的有益效果
有益效果
采用上述技术方案所产生的有益效果在于：
1、本发明相较于传统的深度学习算法，通过数据增强、迁移学习以及对网络 结构进行改进等手段加快了模型训练的收敛速度，提高了肺炎识别的准确率以及模型的泛化能力；
2、本发明也提供了一个基于Django的BS应用，简化用户安装过程，减轻对本地环境的依赖。通过友好的界面以及围绕实际需求而设计的系统模块，可以使得本系统很容易地被各个医院接受，也可以很容易得和医院的其他系统进行功能上的融合。基于MVC架构分的应用也提高了后期维护以及再开发的效率；
3、通过智能辅助诊断系统，也可以大大减轻了人工阅片给医生带来的困扰。
对附图的简要说明
附图说明
图1为本发明一种基于深度学习的X-Ray胸片肺炎智能诊断系统架构图；
图2为本发明一种基于深度学习的X-Ray胸片肺炎智能诊断系统简要业务流程图；
图3为本发明用户登录与注册模块流程图；
图4为本发明图片格式转换模块流程图；
图5位本发明肺炎预测模块流程图；
图6为本发明一种基于深度学习的X-Ray胸片肺炎智能诊断方法流程图；
图7为本发明深度卷积神经网络VGG预测模型结构图；
图8为本发明实施例中用户登录界面图；
图9为本发明实施例中用户注册界面图；
图10为本发明实施例中图片格式转换界面图；
图11为本发明实施例中添加病人信息并上传病人X-Ray胸片并进行预测界面图；
图12为本发明实施例中病人X-Ray胸片肺炎预测结果输出界面。
发明实施例
本发明的实施方式
下面结合附图和实施例，对本发明的具体实施方式作进一步详细描述。以下实施例用于说明本发明，但不用来限制本发明的范围。
一种基于深度学习的X-Ray胸片肺炎智能诊断系统，系统包括三个模块分别为 ：用户登录与注册模块、图片格式转换模块和肺炎预测模块，系统的简要业务流程如图2所示；
其中，用户登录与注册模块提供登录功能、注册功能和重置密码功能，用于对用户提供进入系统的入口，其流程如图3所示；
所述登录功能需要用户在登录界面输入账号和密码并提交给系统，用户登录界面如图8所示，系统后端根据输入的账号，向数据库的用户信息表中查询相应的密码，如果返回结果为空，则说明用户输入的账号不存在，如果返货的结果与用户输入的密码不匹配，则说明用户密码输入错误，只有当用户输入的密码和从数据库中查询的密码相匹配时，系统才会显示相应的跳转界面；
所述注册功能是用户可以输入账号、密码、电话、邮箱地址进行注册，用户注册界面如图9所示，后台会通过JS脚本对这些信息进行合法性判断，如果信息全都合法则在数据库的用户信息表中新增一条数据；
所述重置密码功能是当用户忘记密码时，可以在找回密码页面输入账号和邮箱的验证信息，当邮箱验证信息正确时，系统会允许用户进行密码重置，并修改用户信息表中相应的密码信息。
其中，图片格式转换模块提供选择本地图片以及格式转换功能，图片格式转换界面如图10所示，格式转换功能用于将影像中心的dicom格式数据转换为本系统所需jpeg格式。
其中，肺炎预测模块提供病人信息添加以及图片预测功能，用于将病人信息进行录入并将对应的影像数据添加至病人数据库中，添加病人信息并上传病人X-Ray胸片并进行预测界面如图11所示，然后使用本系统中训练好的预测模型对该病人的数据进行预测分析；
所述病人信息添加是用户通过将患者姓名、性别、年龄、就诊日期录入系统，最后将图片，添加至病人信息中，后台会对这些数据进行JS校验，之后保存至数据库中。
所述图片预测功能是选择一张本地图片，通过系统内训练好的预测模型来完成对患者是否罹患肺炎的预测。
该系统运用B/S架构，将系统分成四层，分别为平台层、支撑层、服务层和应 用层，其架构如图1所示；
其中，应用层包含系统调用接口、Web访问接口和结果可视化接口，与用户端相连接；
服务层包含用户注册、用户认证、病人导入、图片导入、用户登录、格式转换、模型加载和图片预测的用户可操作界面；其中，用户注册、用户登录、用户认证属于登录与注册模块提供的服务；格式转换属于图片格式转换模块提供的服务；病人导入、图片导入、模型加载、图片预测属于肺炎预测模块提供的服务；
支撑层包含基于深度卷积神经网络的分类方法、基于关系型数据库的数据库管理系统、传统图像处理方法、医学图像处理方法；其中基于深度卷积神经网络的分类方法、传统图像处理方法为图片预测提供服务；基于关系型数据库的数据库管理系统为用户注册、用户登录、用户认证提供服务；医学图像处理方法为格式转换提供服务；
平台层采用Keras框架，按照卷积层、BN层、高级激活函数层、池化层、全连接层以及softmax层对深度卷积神经网络进行设计，选用损失函数binary cross-entropy及优化函数RMSProp对卷积神经网络进行优化；采用sqlite3关系型数据库作为本系统的数据库管理系统；采用ITK软件库作为医学图像处理方法的平台；采用opencv计算机视觉库作为传统图像处理方法的平台。
为解决上述技术问题，本发明所采取的技术方案是：一种采用基于深度学习的X-Ray胸片肺炎智能诊断系统进行诊断的方法，该方法的流程如图6所示，包括如下步骤：
步骤1：获取一张dicom格式的X-Ray胸片图像，调用ITK库中的ReadImage方法读取dicom格式图片，之后通过调用ITK中GetArrayFromImage方法提取dicom图像中的像素矩阵，最后通过opencv中的imwrite方法将已经提取的像素矩阵保存成jpeg格式的图片；
步骤2：选择数据集Chest X-Ray Images(Pneumonia)，生成训练集和测试集；
步骤3：建立深度卷积神经网络VGG预测模型，如图7所示，其中VGG模型包括六个卷积层、BN层、高级激活函数，两个全连接层以及最后的softmax层，设 置模型训练迭代次数epoch＝50；
对于模型的权重加载采用ImageNet训练好的权重对模型进行迁移学习，在卷积层和激活函数之间加入BatchNormalization层来加快网络收敛速度，高级激活函数LeakyRelu代替Relu，采用交叉熵损失函数binary cross-entropy来作为模型优化指标，优化函数采用RMSProp方法加快模型收敛；
步骤4：用训练集对VGG预测模型进行训练；
步骤5：采用翻转、旋转、仿射变换这些数据增强的方法，对训练集的图像数据进行增强处理，得到新的训练集；
步骤6：输入测试集，对训练好的VGG预测模型进行测试，得到预测准确率；
步骤7：重复执行步骤4至步骤7，对深度卷积神经网络VGG预测模型进行迭代训练，直到迭代次数达到预设的epoch的值，停止迭代；
步骤8：将测试集上准确率最高的VGG模型进行保存；
步骤9：将步骤1得到的jpeg格式的图片输入到步骤8保存的准确率最高的VGG模型中，得到图片的分类预测结果。
本实施例中设置预测结果输出规则：1为肺炎，0为正常；
本实施例中病人的X-Ray胸片图像预测结果输出为0，可判断该病人没有患肺炎，结果输出界面如图12所示。

Claims (6)

  1. 一种基于深度学习的X-Ray胸片肺炎智能诊断系统，其特征在于包括用户登录与注册模块、图片格式转换模块和肺炎预测模块，该系统运用B/S架构，将系统分成四层，分别为平台层、支撑层、服务层和应用层；
    其中，应用层包含系统调用接口、Web访问接口和结果可视化接口，与用户端相连接；
    服务层包含用户注册、用户认证、病人导入、图片导入、用户登录、格式转换、模型加载和图片预测的用户可操作界面；其中，用户注册、用户登录、用户认证属于登录与注册模块提供的服务；格式转换属于图片格式转换模块提供的服务；病人导入、图片导入、模型加载、图片预测属于肺炎预测模块提供的服务；
    支撑层包含基于深度卷积神经网络的分类方法、基于关系型数据库的数据库管理系统、传统图像处理方法、医学图像处理方法；其中基于深度卷积神经网络的分类方法、传统图像处理方法为图片预测提供服务；基于关系型数据库的数据库管理系统为用户注册、用户登录、用户认证提供服务；医学图像处理方法为格式转换提供服务；
    平台层采用Keras框架，按照卷积层、BN层、高级激活函数层、池化层、全连接层以及softmax层对深度卷积神经网络进行设计，选用损失函数binary cross-entropy及优化函数RMSProp对卷积神经网络进行优化；采用sqlite3关系型数据库作为本系统的数据库管理系统；采用ITK软件库作为医学图像处理方法的平台；采用opencv计算机视觉库作为传统图像处理方法的平台。
  2. 根据权利要求1所述的一种基于深度学习的X-Ray胸片肺炎智能诊断系统，其特征在于用户登录与注册模块提供登录功能、注册功能和重置密码功能，用于对用户提供进入系统的入口；
    所述登录功能需要用户在登录界面输入账号和密码并提交给系统 ，系统后端根据输入的账号，向数据库的用户信息表中查询相应的密码，如果返回结果为空，则说明用户输入的账号不存在，如果返回的结果与用户输入的密码不匹配，则说明用户密码输入错误，只有当用户输入的密码和从数据库中查询的密码相匹配时，系统才会显示相应的跳转界面；
    所述注册功能是用户可以输入账号、密码、电话、邮箱地址进行注册，后台会通过JS脚本对这些信息进行合法性判断，如果信息全都合法则在数据库的用户信息表中新增一条数据；
    所述重置密码功能是当用户忘记密码时，可以在找回密码页面输入账号和邮箱的验证信息，当邮箱验证信息正确时，系统会允许用户进行密码重置，并修改用户信息表中相应的密码信息。
  3. 根据权利要求1所述的一种基于深度学习的X-Ray胸片肺炎智能诊断系统，其特征在于图片格式转换模块提供选择本地图片以及格式转换功能，格式转换功能用于将影像中心的dicom格式数据转换为本系统所需jpeg格式。
  4. 根据权利要求1所述的一种基于深度学习的X-Ray胸片肺炎智能诊断系统，其特征在于肺炎预测模块提供病人信息添加以及图片预测功能，用于将病人信息进行录入并将对应的影像数据添加至病人数据库中，然后使用本系统中训练好的预测模型对该病人的数据进行预测分析；
    所述病人信息添加是用户通过将患者姓名、性别、年龄、就诊日期录入系统，最后将图片添加至病人信息中，后台会对这些数据进行JS校验，之后保存至数据库中；
    所述图片预测功能是选择一张本地图片，通过系统内训练好的预测模型来完成对患者是否罹患肺炎的预测。
  5. 一种采用权利要求1所述的基于深度学习的X-Ray胸片肺炎智能诊断系统进行诊断的方法，其特征在于包括如下步骤：
    步骤1：获取一张dicom格式的X-Ray胸片图像，调用ITK库中的Re adImage方法读取dicom格式图片，之后通过调用ITK中GetArrayFromImage方法提取dicom图像中的像素矩阵，最后通过opencv中的imwrite方法将已经提取的像素矩阵保存成jpeg格式的图片；
    步骤2：选择数据集Chest X-Ray Images(Pneumonia)，生成训练集和测试集；
    步骤3：建立深度卷积神经网络VGG预测模型，其中VGG模型包括六个卷积层、BN层、高级激活函数，两个全连接层以及最后的softmax层，并设置模型训练迭代次数epoch的值；
    步骤4：用训练集对VGG预测模型进行训练；
    步骤5：采用翻转、旋转、仿射变换这些数据增强的方法，对训练集的图像数据进行增强处理，得到新的训练集；
    步骤6：输入测试集，对训练好的VGG预测模型进行测试，得到预测准确率；
    步骤7：重复执行步骤4至步骤7，对深度卷积神经网络VGG预测模型进行迭代训练，直到迭代次数达到预设的epoch的值，停止迭代；
    步骤8：将测试集上准确率最高的VGG模型进行保存；
    步骤9：将步骤1得到的jpeg格式的图片输入到步骤8保存的准确率最高的VGG模型中，得到图片的分类预测结果。
  6. 根据权利要求5所述的一种采用权利要求1所述的基于深度学习的X-Ray胸片肺炎智能诊断系统进行诊断的方法，其特征在于所述步骤3中建立深度卷积神经网络VGG预测模型，对于模型的权重加载采用ImageNet训练好的权重对模型进行迁移学习，在卷积层和激活函数之间加入BatchNormalization层来加快网络收敛速度，高级激活函数LeakyRelu代替Relu，采用交叉熵损失函数binary cross-entropy来作为模型优化指标，优化函数采用RMSProp方法加快模型收敛。
PCT/CN2019/107225 2019-08-20 2019-09-23 一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法 Ceased WO2021031279A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910767278.4 2019-08-20
CN201910767278.4A CN110504027A (zh) 2019-08-20 2019-08-20 一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法

Publications (1)

Publication Number Publication Date
WO2021031279A1 true WO2021031279A1 (zh) 2021-02-25

Family

ID=68588848

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/107225 Ceased WO2021031279A1 (zh) 2019-08-20 2019-09-23 一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法

Country Status (2)

Country Link
CN (1) CN110504027A (zh)
WO (1) WO2021031279A1 (zh)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569866A (zh) * 2021-07-15 2021-10-29 桂林电子科技大学 一种基于深度学习识别hpv试纸的方法
CN113610807A (zh) * 2021-08-09 2021-11-05 西安电子科技大学 基于弱监督多任务学习的新冠肺炎分割方法
CN114461516A (zh) * 2021-12-23 2022-05-10 连邦网络科技服务南通有限公司 一种可数据自动删除的软件图形接口测试方法及服务器
CN114821826A (zh) * 2022-02-03 2022-07-29 北京泽桥医疗科技股份有限公司 一种面向交互式三维人体医学模型的可视化系统
RU2789260C1 (ru) * 2021-12-20 2023-01-31 Автономная некоммерческая организация высшего образования "Университет Иннополис" Система поддержки принятия врачебных решений на основе анализа медицинских изображений
CN116468959A (zh) * 2023-06-15 2023-07-21 清软微视(杭州)科技有限公司 工业缺陷分类方法、装置、电子设备及存储介质
CN116525077A (zh) * 2023-06-29 2023-08-01 安翰科技(武汉)股份有限公司 人工智能医疗器械测试数据的处理方法及系统
CN116561361A (zh) * 2023-05-05 2023-08-08 济南星齐医学检验有限公司 一种用于医学图像阅读行为的数据分析方法和系统
CN118213066A (zh) * 2024-03-08 2024-06-18 上海海神医疗电子仪器有限公司 一种基于dpcnn的腕管综合征智能诊断系统

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112349425A (zh) * 2020-02-10 2021-02-09 胡秋明 新型冠状病毒感染肺炎人工智能快速筛查系统
CN112631898A (zh) * 2020-12-09 2021-04-09 南京理工大学 基于cnn-svm的软件缺陷预测方法
CN113205126A (zh) * 2021-04-27 2021-08-03 武汉工程大学 一种基于神经网络的胸部x光片识别检测系统和方法
CN113476065B (zh) * 2021-09-06 2022-04-22 华南理工大学 一种多类肺炎诊断系统
CN114970372A (zh) * 2022-06-27 2022-08-30 内蒙古大学 基于机器学习算法的数据预测软件系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104346503A (zh) * 2013-07-23 2015-02-11 广州华久信息科技有限公司 一种基于人脸图像的情感健康监控方法及手机
CN104732468A (zh) * 2015-04-08 2015-06-24 南京吉星兆健康信息咨询有限公司 一种基于移动互联的远程医疗系统
CN108776962A (zh) * 2018-04-11 2018-11-09 浙江师范大学 一种构建肺肿瘤良恶性预测模型的方法
CN108876775A (zh) * 2018-06-12 2018-11-23 广州图灵人工智能技术有限公司 糖尿病性视网膜病变的快速检测方法
CN109379329A (zh) * 2018-09-05 2019-02-22 中国人民解放军战略支援部队信息工程大学 基于lstm的网络安全协议模糊测试方法及系统
CN109686444A (zh) * 2018-12-27 2019-04-26 上海联影智能医疗科技有限公司 用于医学图像分级的系统和方法
CN109712693A (zh) * 2018-12-14 2019-05-03 贵州大学 一种智能病理诊断方法及系统
WO2019136909A1 (zh) * 2018-01-12 2019-07-18 平安科技（深圳）有限公司 基于深度学习的语音活体检测方法、服务器及存储介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106372390B (zh) * 2016-08-25 2019-04-02 汤一平 一种基于深度卷积神经网络的预防肺癌自助健康云服务系统
CN109411036A (zh) * 2017-05-09 2019-03-01 北京华信诚达科技有限公司 一种监测各种疾病手术并发症的云平台系统和方法
CN107749308A (zh) * 2017-10-18 2018-03-02 北京航空航天大学 一种为甲状腺超声诊断app提供支撑的后台管理系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104346503A (zh) * 2013-07-23 2015-02-11 广州华久信息科技有限公司 一种基于人脸图像的情感健康监控方法及手机
CN104732468A (zh) * 2015-04-08 2015-06-24 南京吉星兆健康信息咨询有限公司 一种基于移动互联的远程医疗系统
WO2019136909A1 (zh) * 2018-01-12 2019-07-18 平安科技（深圳）有限公司 基于深度学习的语音活体检测方法、服务器及存储介质
CN108776962A (zh) * 2018-04-11 2018-11-09 浙江师范大学 一种构建肺肿瘤良恶性预测模型的方法
CN108876775A (zh) * 2018-06-12 2018-11-23 广州图灵人工智能技术有限公司 糖尿病性视网膜病变的快速检测方法
CN109379329A (zh) * 2018-09-05 2019-02-22 中国人民解放军战略支援部队信息工程大学 基于lstm的网络安全协议模糊测试方法及系统
CN109712693A (zh) * 2018-12-14 2019-05-03 贵州大学 一种智能病理诊断方法及系统
CN109686444A (zh) * 2018-12-27 2019-04-26 上海联影智能医疗科技有限公司 用于医学图像分级的系统和方法

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569866A (zh) * 2021-07-15 2021-10-29 桂林电子科技大学 一种基于深度学习识别hpv试纸的方法
CN113610807B (zh) * 2021-08-09 2024-02-09 西安电子科技大学 基于弱监督多任务学习的新冠肺炎分割方法
CN113610807A (zh) * 2021-08-09 2021-11-05 西安电子科技大学 基于弱监督多任务学习的新冠肺炎分割方法
RU2789260C1 (ru) * 2021-12-20 2023-01-31 Автономная некоммерческая организация высшего образования "Университет Иннополис" Система поддержки принятия врачебных решений на основе анализа медицинских изображений
CN114461516A (zh) * 2021-12-23 2022-05-10 连邦网络科技服务南通有限公司 一种可数据自动删除的软件图形接口测试方法及服务器
CN114461516B (zh) * 2021-12-23 2024-03-29 连邦网络科技服务南通有限公司 一种可数据自动删除的软件图形接口测试方法及服务器
CN114821826A (zh) * 2022-02-03 2022-07-29 北京泽桥医疗科技股份有限公司 一种面向交互式三维人体医学模型的可视化系统
CN116561361A (zh) * 2023-05-05 2023-08-08 济南星齐医学检验有限公司 一种用于医学图像阅读行为的数据分析方法和系统
CN116468959A (zh) * 2023-06-15 2023-07-21 清软微视(杭州)科技有限公司 工业缺陷分类方法、装置、电子设备及存储介质
CN116468959B (zh) * 2023-06-15 2023-09-08 清软微视(杭州)科技有限公司 工业缺陷分类方法、装置、电子设备及存储介质
CN116525077A (zh) * 2023-06-29 2023-08-01 安翰科技(武汉)股份有限公司 人工智能医疗器械测试数据的处理方法及系统
CN116525077B (zh) * 2023-06-29 2023-09-08 安翰科技(武汉)股份有限公司 人工智能医疗器械测试数据的处理方法及系统
CN118213066A (zh) * 2024-03-08 2024-06-18 上海海神医疗电子仪器有限公司 一种基于dpcnn的腕管综合征智能诊断系统

Also Published As

Publication number Publication date
CN110504027A (zh) 2019-11-26

Similar Documents

Publication Publication Date Title
WO2021031279A1 (zh) 一种基于深度学习的X-Ray胸片肺炎智能诊断系统与方法
Choe et al. Content-based image retrieval by using deep learning for interstitial lung disease diagnosis with chest CT
Khan et al. Deep neural architectures for medical image semantic segmentation
Ma et al. Thyroid diagnosis from SPECT images using convolutional neural network with optimization
Wu et al. Development and evaluation of a deep learning algorithm for rib segmentation and fracture detection from multicenter chest CT images
Kaur et al. Methods for automatic generation of radiological reports of chest radiographs: a comprehensive survey
US10692602B1 (en) Structuring free text medical reports with forced taxonomies
Tan et al. Automated vessel segmentation in lung CT and CTA images via deep neural networks
CN116631584A (zh) 通用型医学影像报告生成方法与系统、电子设备及可读存储介质
JP2023114341A (ja) 情報処理装置、情報処理方法、及びプログラム
Wang et al. ICA-Unet: An improved U-net network for brown adipose tissue segmentation
Xue et al. Multi-modal tumor segmentation methods based on deep learning: a narrative review
Song et al. Medical image classification: Knowledge transfer via residual U-Net and vision transformer-based teacher-student model with knowledge distillation
Li et al. Few‐shot learning with deformable convolution for multiscale lesion detection in mammography
Kumar et al. Transformer enabled multi-modal medical diagnosis for tuberculosis classification
A Hasib et al. YOLOv8 framework for COVID-19 and pneumonia detection using synthetic image augmentation
Jiang et al. Fusion of medical imaging and electronic health records with attention and multi-head machanisms
Deng et al. A diagnostic report supervised deep learning model training strategy for diagnosis of COVID-19
WO2019208130A1 (ja) 医療文書作成支援装置、方法およびプログラム、学習済みモデル、並びに学習装置、方法およびプログラム
Zhou et al. An adaptive and lightweight YOLOv5 detection model for lung tumor in PET/CT images
CN114359194A (zh) 基于改进U-Net网络的多模态脑卒中梗死区域图像处理方法
Souza et al. Advanced breast cancer detection using spatial attention and neural architecture search (SANAS-Net)
Hassan et al. Analysis of multimodal representation learning across medical images and reports using multiple vision and language pre-trained models
Liu et al. Detection of fetal facial anatomy in standard ultrasonographic sections based on real‐time target detection network
Baaklini et al. Deep learning for MRI-based acute and subacute ischaemic stroke lesion segmentation—a systematic review, meta-analysis, and pilot evaluation of key results

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19942440

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19942440

Country of ref document: EP

Kind code of ref document: A1