WO2019052063A1 - 基于人工智能的医学影像分类处理系统及方法 - Google Patents

基于人工智能的医学影像分类处理系统及方法 Download PDF

Info

Publication number
WO2019052063A1
WO2019052063A1 PCT/CN2017/116664 CN2017116664W WO2019052063A1 WO 2019052063 A1 WO2019052063 A1 WO 2019052063A1 CN 2017116664 W CN2017116664 W CN 2017116664W WO 2019052063 A1 WO2019052063 A1 WO 2019052063A1
Authority
WO
WIPO (PCT)
Prior art keywords
medical
feature
image
medical image
training
Prior art date
Application number
PCT/CN2017/116664
Other languages
English (en)
French (fr)
Inventor
姚育东
钱唯
郑斌
马贺
齐守良
赵明芳
Original Assignee
深圳市前海安测信息技术有限公司
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 深圳市前海安测信息技术有限公司 filed Critical 深圳市前海安测信息技术有限公司
Publication of WO2019052063A1 publication Critical patent/WO2019052063A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques

Definitions

  • the present invention relates to the field of medical image processing technology, and in particular, to a medical image classification processing system and method based on artificial intelligence.
  • Medical institutions produce a huge amount of medical images, and image data often contains a large amount of potential information.
  • the medical market relies mainly on manual interpretation of medical images, which is inefficient and has limited information to be tapped, and cannot fully utilize data resources.
  • computer learning in various fields, including computer vision has been rapidly occupied by deep learning. Deep learning has achieved remarkable results in the fields of image recognition and speech recognition.
  • its powerful automatic feature extraction complex model construction and image processing capabilities are very suitable for dealing with new problems faced by medical image data analysis, which has aroused widespread concern in the field of biomedical researchers.
  • medical image-assisted processing cannot establish a medical model based on the feature classification training of medical images in different parts of the human body, which affects the efficiency and accuracy of medical examinations performed by doctors.
  • the main object of the present invention is to provide a medical image classification processing system and method based on artificial intelligence, which aims to solve the problem that the existing medical image auxiliary processing cannot affect the medical examination accuracy according to the characteristic classification training of medical images in different parts of the human body. problem.
  • the present invention provides a medical image classification processing system based on artificial intelligence, which is applied to a server, which is connected with a medical image information library, a medical image acquisition terminal and a doctor workstation.
  • the medical image classification processing system includes: an image information acquisition module, configured to obtain medical image information from a medical image information library; a video information classification module, configured to classify medical image information based on different human body parts; a medical model building module, The utility model is used for establishing a medical evaluation model corresponding to various parts of the human body based on the classified medical image information; the medical model training module is used for training the medical evaluation model of various parts of the human body, and extracting medical model parameters from the trained medical evaluation model And storing the medical model parameters in the storage unit; the medical image processing module is configured to receive the medical examination image of the patient from the medical image acquisition terminal, identify the examination site of the patient from the medical examination image of the patient, and check according to the patient The part obtains the corresponding medical model parameter from the storage unit; the medical image output module is configured
  • the medical model training module includes a feature training sub-module and a model fine-tuning sub-module, wherein: the feature training sub-module is configured to extract local feature information from the medical evaluation model, and use an artificial neural network to local features The information is subjected to feature training to obtain a set of feature vectors; the model fine-tuning sub-module is used to apply the feature vector to the training of the convolutional neural network, and fine-tune the medical evaluation model according to the training result.
  • the feature training sub-module is configured to extract local feature information from the medical evaluation model, and use an artificial neural network to local features The information is subjected to feature training to obtain a set of feature vectors
  • the model fine-tuning sub-module is used to apply the feature vector to the training of the convolutional neural network, and fine-tune the medical evaluation model according to the training result.
  • the artificial neural network is a layered neural network implemented by an automatic encoder, which uses local feature information as an input of the first layer and performs feature training, and uses the output of the first layer as the second layer. Input and feature training, and the output of the second layer as the input of the third layer and feature training.
  • the convolutional neural network comprises a plurality of neurons, a feature extraction layer and a feature mapping layer, wherein each neuron is connected to a feature mapping layer, and the feature mapping layer extracts feature information from each neuron.
  • the feature mapping layer is composed of a plurality of feature vector maps.
  • the medical examination image comprises a nuclear magnetic image, a CT image, an ultrasound image, an X-ray image, and an infrared image.
  • the invention also provides a medical image classification processing method based on artificial intelligence, which is applied to a server, which is connected with a medical image information library, a medical image acquisition terminal and a doctor workstation, and the medical image classification processing method comprises the steps of: from medicine Obtaining medical image information in the image information base; classifying and processing the medical image information based on different human body parts; establishing a medical evaluation model corresponding to each part of the human body based on the classified medical image information; training the medical evaluation model of each part of the human body; The medical model is extracted from the trained medical evaluation model, and the medical model parameters are saved in the storage unit of the server; the medical examination image of the patient is received from the medical image acquisition terminal; and the examination site of the patient is identified from the medical examination image of the patient And obtaining corresponding medical model parameters from the storage unit according to the examination site of the patient; sending the medical examination image of the patient and the corresponding medical model parameters to the doctor workstation for medical examination by the doctor.
  • the step of performing training processing on the medical evaluation model of each part of the human body comprises: extracting local feature information from the medical evaluation model, and performing feature training on the extracted local feature information by using an artificial neural network to obtain a group Feature vector; applying the feature vector to the training of the convolutional neural network, and fine-tuning the medical evaluation model according to the training result.
  • the artificial neural network is a layered neural network implemented by an automatic encoder, which uses local feature information as an input of the first layer and performs feature training, and uses the output of the first layer as the second layer. Input and feature training, and the output of the second layer as the input of the third layer and feature training.
  • the convolutional neural network comprises a plurality of neurons, a feature extraction layer and a feature mapping layer, wherein each neuron is connected to a feature mapping layer, and the feature mapping layer extracts feature information from each neuron.
  • the feature mapping layer is composed of a plurality of feature vector maps.
  • the medical examination image comprises a nuclear magnetic image, a CT image, an ultrasound image, an X-ray image, and an infrared image.
  • the artificial intelligence-based medical image classification processing system and method of the present invention adopts the above technical solutions, and achieves the following technical effects: obtaining medical image information from a medical image information database, and based on different parts The medical image information classification processing is performed, and then based on the classified medical image information, a medical evaluation model corresponding to each human body part is established, the medical evaluation model is trained, and the model parameters are saved as a follow-up doctor's human body image examination as a reference basis. , assisting doctors to improve the efficiency and accuracy of medical examination of patients.
  • FIG. 1 is a schematic diagram of an application environment of a preferred embodiment of a medical image classification processing system based on artificial intelligence according to the present invention
  • FIG. 2 is a flow chart of a preferred embodiment of a medical image classification processing method based on artificial intelligence according to the present invention.
  • FIG. 1 is a schematic diagram of an application environment of a preferred embodiment of a medical image classification processing system based on artificial intelligence according to the present invention.
  • the medical image classification processing system 10 is installed and operated in the server 1.
  • the server 1 establishes a data connection with the medical image information repository 2 via a database connection, and establishes a communication connection with the medical image acquisition terminal 3 and the doctor workstation 4 via the communication network 5.
  • the server 1 can be a cloud computing device with data processing and communication functions, such as a large computer or a server
  • the database connection can be an open database connection (Open Database Connectivity (ODBC) and Java Database Connectivity (JDBC).
  • ODBC Open Database Connectivity
  • JDBC Java Database Connectivity
  • the medical image information base 2 stores different groups of human medical image information.
  • the different groups in the present invention can be divided according to gender, for example, male or female groups, and can also be divided according to different age groups, such as infants and children, children, Teenagers, adolescents, middle-aged, elderly and
  • the medical image capturing terminal 3 is a medical detecting device installed in a medical examination institution such as a community medical workstation or a hospital, such as a nuclear magnetic, CT, ultrasonic, X-ray machine, infrared instrument, or the like.
  • the medical image capturing terminal 3 is configured to collect medical examination images of the patient, for example, nuclear magnetic images, CT images, ultrasound images, X-ray images, and infrared images, and send the medical examination images of the patients to the server 1 through the communication network 3.
  • the doctor workstation 4 is a doctor workstation computer installed in a medical examination institution (for example, a second and third hospital, a social health center) for displaying a medical examination image of the patient and corresponding medical model parameters, and is available for the doctor to perform medical examination according to the medical examination.
  • the image and its corresponding medical model parameters are medically examined by the patient.
  • the communication network 5 may be an internet network including a local area network (LAN), a wide area network (WAN), or a wireless transmission network including GSM, GPRS, and CDMA for transmission of medical image information.
  • the server 1 includes, but is not limited to, a medical image classification processing system 10, a storage unit 11, a processing unit 12, and a communication unit 13.
  • the storage unit 11 and the communication unit 13 are both connected to the processing unit 12 via a data bus, and can perform information interaction with the medical image classification processing system 10 through the processing unit 12.
  • the storage unit 11 can be a read only memory unit ROM, an electrically erasable memory unit EEPROM or a flash memory unit FLASH.
  • the processing unit 12 can be a central processing unit (CPU), a microprocessor, a microcontroller (MCU), a data processing chip, or an information processing unit having data processing functions.
  • the communication unit 13 can be a wireless communication interface with remote wireless communication functions, such as a communication interface supporting GSM, GPRS, and CDMA.
  • the medical image classification processing system 10 includes, but is not limited to, an image information acquisition module 101, a video information classification module 102, a medical model establishment module 103, a medical model training module 104, and a medical image processing module 105. And a medical image output module 106.
  • the medical model training module 104 includes a feature training sub-module 1041 and a model fine-tuning sub-module 1042.
  • the module referred to in the present invention refers to a series of computer program instruction segments that can be executed by the processing unit 12 of the server 1 and that can perform a fixed function, which are stored in the storage unit 11 of the server 1.
  • the image information acquiring module 101 is configured to obtain medical image information from the medical image information base 2.
  • the medical image information includes, but is not limited to, medical examination images generated by medical testing equipment such as nuclear magnetic, CT, ultrasonic, X-ray machine, infrared instrument, etc., and the medical inspection images can be digitally adopted through the DICOM3.0 international standard interface. Medical evaluation model input image information.
  • the image information classification module 102 is configured to classify the medical image information based on different human body parts. Due to the non-uniformity of the human body part data, the invention classifies the medical image information of each part, and takes the medical image information of the same part as a sub-class, which is convenient for the subsequent medical evaluation model calculation and processing.
  • the medical model establishing module 103 is configured to establish a medical evaluation model corresponding to each part of the human body based on the classified medical image information.
  • the establishment of each medical evaluation model corresponding to each part of the human body is based on the classified medical image information, so each medical evaluation model is independent of each other and can be learned in parallel.
  • the medical model training module 104 is configured to perform training processing on medical evaluation models of various parts of the human body, extract medical model parameters from the trained medical evaluation model, and save the medical model parameters in the storage unit 11.
  • the invention establishes and calculates medical model parameters based on the medical evaluation model of the classified data, and provides a reference for the follow-up doctor to perform medical examination for the patient.
  • the feature training sub-module 1041 is configured to extract local feature information from the medical evaluation model and utilize an artificial neural network (Artificial) The Neural Network (ANN) performs feature training on the extracted local feature information to obtain a set of feature vectors.
  • the artificial neural network is a hierarchical neural network implemented by an automatic encoder, which uses local feature information as an input of the first layer and performs feature training, and uses the output of the first layer as The second layer of input and feature training, the output of the second layer as the input of the third layer and feature training.
  • the automatic encoder takes the extracted medical image information as an input, performs feature training through the coding of the first layer to the second layer, and restores the medical image information through the decoding of the second layer to the third layer.
  • the model fine-tuning sub-module 1042 is configured to apply the feature vector to the training of the convolutional neural network, and fine-tune the medical evaluation model according to the training result.
  • the Convolutional Neural Network is a feedforward neural network.
  • the convolutional neural network includes a plurality of neurons, a feature extraction layer, and a feature mapping layer, wherein each neuron is connected to a feature mapping layer, and the feature mapping layer extracts feature information from each neuron.
  • the feature mapping layer is composed of a plurality of feature vector maps. Each feature vector map is a plane, and all neurons on the plane have equal weights.
  • Each convolutional layer in the convolutional neural network is followed by a computational layer for local averaging and quadratic extraction.
  • This unique two-feature extraction structure reduces feature resolution.
  • the invention combines unsupervised feature training and convolutional neural network in deep learning, can not only apply medical image information of more medical evaluation models, but also enhance the feature training ability of convolutional neural networks and reduce the difficulty of medical evaluation model training. .
  • the medical image processing module 105 is configured to receive a medical examination image of the patient from the medical image acquisition terminal 3, identify the examination site of the patient from the medical examination image of the patient, and acquire the examination site from the storage unit 11 according to the examination site of the patient. Corresponding medical model parameters. In this embodiment, if the medical image capturing terminal 3 collects the lung image of the patient, the medical image processing module 105 can recognize the human body examination site as the lung according to the structural characteristics of the human body, and acquire the patient lung from the storage unit 11. Corresponding medical model parameters.
  • the medical image output module 106 is configured to send the medical examination image of the patient and the corresponding medical model parameters to the doctor workstation 4 for medical examination by the doctor, and can assist the doctor to accurately analyze the medical examination image and make medicine. Check the report to assist the doctor in improving the efficiency and accuracy of the medical examination of the patient.
  • FIG. 2 is a flow chart of a preferred embodiment of the medical image classification processing method based on artificial intelligence according to the present invention.
  • the artificial intelligence-based medical image classification processing method is applied to the server 1 , and the artificial intelligence-based medical image classification processing method includes the following steps S21 to S28 .
  • the image information acquiring module 101 acquires medical image information from the medical image information base 2.
  • the medical image information includes, but is not limited to, medical examination images generated by medical testing equipment such as nuclear magnetic, CT, ultrasonic, X-ray machine, infrared instrument, etc., and the medical examination images can pass the DICOM3.0 international standard.
  • the interface is digitally used as image information for medical evaluation models.
  • step S22 the image information classification module 102 classifies the medical image information based on different human body parts. Due to the non-uniformity of the human body part data, the invention classifies the medical image information of each part, and takes the medical image information of the same part as a sub-class, which is convenient for the subsequent medical evaluation model calculation and processing.
  • step S23 the medical model establishing module 103 establishes a medical evaluation model corresponding to each part of the human body based on the classified medical image information.
  • the establishment of each medical evaluation model corresponding to each part of the human body is based on the classified medical image information, so each medical evaluation model is independent of each other and can be learned in parallel.
  • the medical model training module 104 performs training processing on the medical evaluation model of each part of the human body.
  • the training process for the medical evaluation model of each part of the human body comprises the steps of: extracting local feature information from the medical evaluation model, and performing feature training on the extracted local feature information by using an artificial neural network to obtain a set of feature vectors;
  • the feature vector is applied to the training of the convolutional neural network, and the medical evaluation model is fine-tuned according to the training result.
  • the invention combines unsupervised feature training and convolutional neural network in deep learning, can not only apply medical image information of more medical evaluation models, but also enhance the feature training ability of convolutional neural networks and reduce the difficulty of medical evaluation model training. .
  • step S25 the medical model training module 104 extracts medical model parameters from the trained medical evaluation model, and saves the medical model parameters in the storage unit 11.
  • the invention establishes and calculates medical model parameters based on the medical evaluation model of the classified data, and provides a reference for the follow-up doctor to perform medical examination for the patient.
  • the medical image processing module 105 receives the medical examination image of the patient from the medical image acquisition terminal 3.
  • the medical image capturing terminal 3 may be a medical detecting device such as a nuclear magnetic, CT, ultrasonic, X-ray machine, or infrared instrument.
  • the medical image capturing terminal 3 collects medical examination images of the patient, for example, nuclear magnetic images, CT images, ultrasound images, X-ray images, and infrared images, and transmits the medical examination images of the patients to the server 1 through the communication network 3.
  • the medical image processing module 105 identifies the examination site of the patient from the medical examination image of the patient, and acquires the medical model parameter corresponding to the examination site from the storage unit 11 according to the examination site of the patient.
  • the medical image processing module 105 can recognize the human body examination site as the lung according to the structural characteristics of the human body, and acquire the patient lung from the storage unit 11. Corresponding medical model parameters.
  • Step S28 the medical image output module 106 sends the medical examination image of the patient and the corresponding medical model parameters to the doctor workstation 4 for the medical examination of the patient, which can assist the doctor to accurately analyze the medical examination image and make a medical examination. Reports to assist doctors in improving the efficiency and accuracy of medical examinations of patients.
  • the artificial intelligence-based medical image classification processing system and method of the present invention acquires medical image information from a medical image information database, classifies the medical image information based on different parts, and then establishes a correspondence based on the classified medical image information.
  • a medical evaluation model of each human body part training and processing the medical evaluation model, and finally saving the model parameters as a reference for the follow-up doctor to examine the human body image, and assisting the doctor to improve the efficiency and accuracy of medical examination of the patient.
  • the artificial intelligence-based medical image classification processing system and method of the present invention adopts the above technical solutions, and achieves the following technical effects: obtaining medical image information from a medical image information database, and based on different parts The medical image information classification processing is performed, and then based on the classified medical image information, a medical evaluation model corresponding to each human body part is established, the medical evaluation model is trained, and the model parameters are saved as a follow-up doctor's human body image examination as a reference basis. , assisting doctors to improve the efficiency and accuracy of medical examination of patients.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

一种基于人工智能的医学影像分类处理系统及方法，该方法包括步骤：从医学影像信息库中获取医学影像信息（S21）；基于不同人体部位将医学影像信息分类处理（S22）；基于分类后的医学影像信息建立人体各部位的医学评估模型（S23）；对人体各部位的医学评估模型进行训练处理（S24）并保存医学模型参数（S25）；从医学影像采集终端接收患者的医学检查影像（S26）；从患者的医学检查影像中识别出检查部位，并根据检查部位从存储单元中获取对应的医学模型参数（S27）；将患者的医学检查影像及对应的医学模型参数发送到医生工作站（S28）。能够根据人体各个不同部位对医学影像进行特征分类并训练医学模型供医生进行医学检查参考，有利于提高医学检查的效率和准确性。

Description

基于人工智能的医学影像分类处理系统及方法 技术领域
本发明涉及医学影像处理技术领域，尤其涉及一种基于人工智能的医学影像分类处理系统及方法。
 背景技术
医疗机构的医学图像产出数据量十分庞大，而图像数据往往包含大量潜在信息。目前医疗市场主要依靠人工判读分析医学图像，效率较低且能挖掘的信息有限，无法充分利用数据资源。近年来，随着深度学习的快速发展，机器学习各领域包括计算机视觉迅速被深度学习占领，深度学习已经在图像识别、语音识别等领域取得了令人瞩目的成果。作为机器学习中的重要方法之一，其强大的自动特征提取复杂模型构建以及图像处理能力，非常适合处理医学影像数据分析所面临的新问题，引起了生物医学领域研究人员的广泛关注。目前，医学图像辅助处理不能根据人体各个不同部位对医学影像进行特征分类训练建立医学模型，影响医生对患者进行医学检查的效率和准确性。
 技术问题
本发明的主要目的在于提供一种基于人工智能的医学影像分类处理系统及方法，旨在解决现有医学图像辅助处理不能根据人体各个不同部位对医学影像进行特征分类训练而影响医学检查准确性的问题。
 技术解决方案
为实现上述目的，本发明提供了一种基于人工智能的医学影像分类处理系统，应用于服务器中，该服务器连接有医学影像信息库、医学影像采集终端和医生工作站。所述医学影像分类处理系统包括：影像信息获取模块，用于从医学影像信息库中获取医学影像信息；影像信息分类模块，用于基于不同人体部位将医学影像信息分类处理；医学模型建立模块，用于基于分类后的医学影像信息建立对应人体各部位的医学评估模型；医学模型训练模块，用于对人体各部位的医学评估模型进行训练处理，从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在存储单元中；医学影像处理模块，用于从医学影像采集终端接收患者的医学检查影像，从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元中获取对应的医学模型参数；医学影像输出模块，用于将患者的医学检查影像及对应的医学模型参数发送到医生工作站供医生对患者进行医学检查。
优选的，所述医学模型训练模块包括特征训练子模块以及模型微调子模块，其中：所述特征训练子模块用于从所述医学评估模型中抽取局部特征信息，并利用人工神经网络对局部特征信息进行特征训练来获得一组特征向量；所述模型微调子模块用于将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。
优选的，所述人工神经网络是通过自动编码器来实现的分层神经网络，该自动编码器将局部特征信息作为第一层的输入并进行特征训练，将第一层的输出作为第二层的输入并进行特征训练，以及将第二层的输出作为第三层的输入并进行特征训练。
优选的，所述卷积神经网络包括多个神经元、特征提取层以及特征映射层，其中，每个神经元与特征映射层相连，所述特征映射层从每个神经元中提取特征信息，所述特征映射层由多个特征向量映射组成。
优选的，所述医学检查影像包括核磁影像、CT影像、超声影像、X光影像以及红外影像。
本发明还提供一种基于人工智能的医学影像分类处理方法，应用于服务器中，该服务器连接有医学影像信息库、医学影像采集终端和医生工作站，所述医学影像分类处理方法包括步骤：从医学影像信息库中获取医学影像信息；基于不同人体部位将医学影像信息分类处理；基于分类后的医学影像信息建立对应人体各部位的医学评估模型；对人体各部位的医学评估模型进行训练处理；从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在服务器的存储单元中；从医学影像采集终端接收患者的医学检查影像；从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元中获取对应的医学模型参数；将患者的医学检查影像及对应的医学模型参数发送到医生工作站供医生对患者进行医学检查。
优选的，所述对人体各部位的医学评估模型进行训练处理的步骤包括：从所述医学评估模型中抽取局部特征信息，并利用人工神经网络对抽取的局部特征信息进行特征训练来获得一组特征向量；将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。
优选的，所述人工神经网络是通过自动编码器来实现的分层神经网络，该自动编码器将局部特征信息作为第一层的输入并进行特征训练，将第一层的输出作为第二层的输入并进行特征训练，以及将第二层的输出作为第三层的输入并进行特征训练。
优选的，所述卷积神经网络包括多个神经元、特征提取层以及特征映射层，其中，每个神经元与特征映射层相连，所述特征映射层从每个神经元中提取特征信息，所述特征映射层由多个特征向量映射组成。
优选的，所述医学检查影像包括核磁影像、CT影像、超声影像、X光影像以及红外影像。
 有益效果
相较于现有技术，本发明所述基于人工智能的医学影像分类处理系统及方法采用上述技术方案，达到了如下技术效果：通过从医学影像信息库获取医学影像信息，并基于不同部位将所述医学影像信息分类处理，然后基于分类后的医学影像信息，建立对应各人体部位的医学评估模型，对所述医学评估模型训练处理，保存模型参数作为后续医生为患者的人体影像检查作为参考依据，辅助医生提高对患者进行医学检查的效率及准确性。
 附图说明
图1是本发明基于人工智能的医学影像分类处理系统优选实施例的应用环境示意图；
图2是本发明基于人工智能的医学影像分类处理方法优选实施例的流程图。
本发明目的实现、功能特点及优点将结合实施例，参照附图做详细说明。
 本发明的最佳实施方式
为更进一步阐述本发明为达成上述目的所采取的技术手段及功效，以下结合附图及较佳实施例，对本发明的具体实施方式、结构、特征及其功效进行详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
参照图1所示，图1是本发明基于人工智能的医学影像分类处理系统优选实施例的应用环境示意图。在本实施例中，所述医学影像分类处理系统10安装并运行于服务器1中。所述服务器1通过数据库连接与医学影像信息库2建立数据连接，以及通过通信网络5与医学影像采集终端3和医生工作站4建立通信连接。所述服务器1可以为一种大型计算机、服务器等具有数据处理和通信功能的云端计算装置，所述数据库连接可以为开放数据库连接（Open Database Connectivity，ODBC）以及Java数据库连接（Java Data Base Connectivity，JDBC）。所述医学影像信息库2存储有不同群体的人体医学影像信息，本发明所称不同群体可以按照性别来划分，例如男、女人群，也可以按照不同年龄段来划分，例如婴幼儿、儿童、少年、青少年、中年、老年人等群体。
所述医学影像采集终端3是设置在社区医疗工作站或者医院等医疗检查机构内的各种医疗检测设备，例如核磁、CT、超声、X光机、红外仪等医疗检测设备。所述医学影像采集终端3用于采集患者的医学检查影像，例如，核磁影像、CT影像、超声影像、X光影像以及红外影像，并将患者的医学检查影像通过通信网络3发送至服务器1。所述医生工作站4为设置在医疗检查机构（例如二三甲医院、社康医疗中心）的医生工作站计算机，用于显示患者的医学检查影像及其对应的医学模型参数，可供医生根据医学检查影像及其对应的医学模型参数对患者进行医学检查。所述通信网络5可以是一种包括局域网（LAN）、广域网（WAN）的网际网络，或者是一种包括GSM、GPRS、CDMA的无线传输网络，用于医学影像信息的传输。
在本实施例中，所述服务器1包括，但不仅限于，医学影像分类处理系统10、存储单元11、处理单元12以及通信单元13。所述存储单元11以及通信单元13均通过数据总线连接至处理单元12，并能通过处理单元12与所述医学影像分类处理系统10进行信息交互。所述存储单元11可以为一种只读存储单元ROM，电可擦写存储单元EEPROM或快闪存储单元FLASH等存储器。所述处理单元12可以为一种中央处理器（CPU）、微处理器、微控制器（MCU）、数据处理芯片、或者具有数据处理功能的信息处理单元。所述通信单元13可以为一种具有远程无线通讯功能的无线通讯接口，例如支持GSM、GPRS、CDMA的通讯接口。
在本实施例中，所述医学影像分类处理系统10包括，但不局限于，影像信息获取模块101、影像信息分类模块102、医学模型建立模块103、医学模型训练模块104、医学影像处理模块105以及医学影像输出模块106。其中，所述医学模型训练模块104包括特征训练子模块1041和模型微调子模块1042。本发明所称的模块是指一种能够被所述服务器1的处理单元12执行并且能够完成固定功能的一系列计算机程序指令段，其存储在所述服务器1的存储单元11中。
所述影像信息获取模块101用于从医学影像信息库2中获取医学影像信息。所述医学影像信息包括，但不限于，核磁、CT、超声、X光机、红外仪等医疗检测设备产生的医学检查影像，这些医学检查影像可以通过DICOM3.0国际标准接口以数字化的方式作为医学评估模型输入的影像信息。
所述影像信息分类模块102用于基于不同人体部位将所述医学影像信息进行分类处理。由于人体部位数据的非统一性，本发明将各部位的医学影像信息进行分类处理，将同一部位的医学影像信息作为一小类，便于后续的医学评估模型计算与处理。
所述医学模型建立模块103用于基于分类后的医学影像信息建立对应人体各部位的医学评估模型。在本实施例中，人体各部位对应的各个医学评估模型的建立是基于分类后的医学影像信息，所以各个医学评估模型之间是相互独立的，可以并行学习训练。
所述医学模型训练模块104用于对人体各部位的医学评估模型进行训练处理，从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在存储单元11中。本发明基于分类数据的医学评估模型建立与计算医学模型参数，为后续医生为患者进行医学检查提供参考依据。
所述特征训练子模块1041用于从所述医学评估模型中抽取局部特征信息，并利用人工神经网络（Artificial Neural Network，ANN）对抽取的局部特征信息进行特征训练来获得一组特征向量。在本实施例中，所述人工神经网络是通过自动编码器来实现的分层神经网络，该自动编码器将局部特征信息作为第一层的输入并进行特征训练，将第一层的输出作为第二层的输入并进行特征训练，将第二层的输出作为第三层的输入并进行特征训练。自动编码器将抽取的医学影像信息作为输入，通过第一层到第二层的编码来实现特征训练，以及通过第二层到第三层的解码来还原医学影像信息。
所述模型微调子模块1042用于将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。在本实施例中，所述卷积神经网络(Convolutional Neural Network，CNN)是一种前馈神经网络。一般地，所述卷积神经网络包括多个神经元、特征提取层以及特征映射层，其中，每个神经元与特征映射层相连，所述特征映射层从每个神经元中提取特征信息，所述特征映射层由多个特征向量映射组成。每个特征向量映射是一个平面，平面上所有神经元的权值相等。卷积神经网络中的每一个卷积层都紧跟着一个用来求局部平均与二次提取的计算层，这种特有的两次特征提取结构减小了特征分辨率。本发明对将深度学习中无监督特征训练与卷积神经网络相结合，不仅能够应用更多医学评估模型的医学影像信息，而且能够增强卷积神经网络的特征训练能力，降低医学评估模型训练难度。
所述医学影像处理模块105用于从医学影像采集终端3接收患者的医学检查影像，从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元11中获取检查部位对应的医学模型参数。在本实施例中，假如医学影像采集终端3采集了患者的肺部影像，医学影像处理模块105根据人体器官结构特征可以识别出人体检查部位为肺部，并从存储单元11中获取患者肺部对应的医学模型参数。
所述医学影像输出模块106用于将患者的医学检查影像及其对应的医学模型参数发送至医生工作站4供医生对患者进行医学检查，可以辅助医生准确地对医学检查影像进行分析并做出医学检查报告，从而辅助医生提高对患者进行医学检查的效率及准确性。
如图2所示，图2是本发明基于人工智能的医学影像分类处理方法优选实施例的流程图。本实施例一并结合图1所示，所述基于人工智能的医学影像分类处理方法应用于服务器1中，该基于人工智能的医学影像分类处理方法包括如下步骤S21至S28。
步骤S21，影像信息获取模块101从医学影像信息库2中获取医学影像信息。在本实施例中，所述医学影像信息包括，但不限于，核磁、CT、超声、X光机、红外仪等医疗检测设备产生的医学检查影像，这些医学检查影像可以通过DICOM3.0国际标准接口以数字化的方式作为医学评估模型输入的影像信息。
步骤S22，影像信息分类模块102基于不同人体部位将所述医学影像信息进行分类处理。由于人体部位数据的非统一性，本发明将各部位的医学影像信息进行分类处理，将同一部位的医学影像信息作为一小类，便于后续的医学评估模型计算与处理。
步骤S23，医学模型建立模块103基于分类后的医学影像信息建立对应人体各部位的医学评估模型。在本实施例中，人体各部位对应的各个医学评估模型的建立是基于分类后的医学影像信息，所以各个医学评估模型之间是相互独立的，可以并行学习训练。
步骤S24，医学模型训练模块104对人体各部位的医学评估模型进行训练处理。所述对人体各部位的医学评估模型进行训练处理包括如下步骤：从所述医学评估模型中抽取局部特征信息，并利用人工神经网络对抽取的局部特征信息进行特征训练来获得一组特征向量；将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。本发明对将深度学习中无监督特征训练与卷积神经网络相结合，不仅能够应用更多医学评估模型的医学影像信息，而且能够增强卷积神经网络的特征训练能力，降低医学评估模型训练难度。
步骤S25，医学模型训练模块104从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在存储单元11中。本发明基于分类数据的医学评估模型建立与计算医学模型参数，为后续医生为患者进行医学检查提供参考依据。
步骤S26，医学影像处理模块105从医学影像采集终端3接收患者的医学检查影像。在本实施例中，所述医学影像采集终端3可以为核磁、CT、超声、X光机、红外仪等医疗检测设备。所述医学影像采集终端3采集患者的医学检查影像，例如，核磁影像、CT影像、超声影像、X光影像以及红外影像，并将患者的医学检查影像通过通信网络3发送至服务器1。
步骤S27，医学影像处理模块105从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元11中获取检查部位对应的医学模型参数。在本实施例中，假如医学影像采集终端3采集了患者的肺部影像，医学影像处理模块105根据人体器官结构特征可以识别出人体检查部位为肺部，并从存储单元11中获取患者肺部对应的医学模型参数。
步骤S28，医学影像输出模块106将患者的医学检查影像及其对应的医学模型参数发送至医生工作站4供医生对患者进行医学检查，可以辅助医生准确地对医学检查影像进行分析并做出医学检查报告，从而辅助医生提高对患者进行医学检查的效率及准确性。
本发明所述基于人工智能的医学影像分类处理系统及方法通过从医学影像信息库获取医学影像信息，并基于不同部位将所述医学影像信息分类处理，然后基于分类后的医学影像信息，建立对应各人体部位的医学评估模型，对所述医学评估模型训练处理，最后保存模型参数作为后续医生为患者的人体影像检查作为参考依据，辅助医生提高对患者进行医学检查的效率及准确性。
本领域技术人员可以理解，上述实施方式中各种方法的全部或部分步骤可以通过相关程序指令完成，该程序可以存储于计算机可读存储介质中，存储介质可以包括：只读存储器、随机存储器、磁盘或光盘等。
以上仅为本发明的优选实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效功能变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。
 工业实用性
相较于现有技术，本发明所述基于人工智能的医学影像分类处理系统及方法采用上述技术方案，达到了如下技术效果：通过从医学影像信息库获取医学影像信息，并基于不同部位将所述医学影像信息分类处理，然后基于分类后的医学影像信息，建立对应各人体部位的医学评估模型，对所述医学评估模型训练处理，保存模型参数作为后续医生为患者的人体影像检查作为参考依据，辅助医生提高对患者进行医学检查的效率及准确性。

Claims (10)

  1. 一种基于人工智能的医学影像分类处理系统，应用于服务器中，该服务器连接有医学影像信息库、医学影像采集终端和医生工作站，其特征在于，所述基于人工智能的医学影像分类处理系统包括：影像信息获取模块，用于从医学影像信息库中获取医学影像信息；影像信息分类模块，用于基于不同人体部位将医学影像信息分类处理；医学模型建立模块，用于基于分类后的医学影像信息建立对应人体各部位的医学评估模型；医学模型训练模块，用于对人体各部位的医学评估模型进行训练处理，从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在服务器的存储单元中；医学影像处理模块，用于从医学影像采集终端接收患者的医学检查影像，从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元中获取对应的医学模型参数；医学影像输出模块，用于将患者的医学检查影像及对应的医学模型参数发送到医生工作站供医生对患者进行医学检查。
  2. 如权利要求1所述的基于人工智能的医学影像分类处理系统，其特征在于，所述医学模型训练模块包括特征训练子模块和模型微调子模块，其中：所述特征训练子模块用于从所述医学评估模型中抽取局部特征信息，并利用人工神经网络对局部特征信息进行特征训练来获得一组特征向量；所述模型微调子模块用于将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。
  3. 如权利要求2所述的基于人工智能的医学影像分类处理系统，其特征在于，所述人工神经网络是通过自动编码器来实现的分层神经网络，该自动编码器将局部特征信息作为第一层的输入并进行特征训练，将第一层的输出作为第二层的输入并进行特征训练，以及将第二层的输出作为第三层的输入并进行特征训练。
  4. 如权利要求2所述的基于人工智能的医学影像分类处理系统，其特征在于，所述卷积神经网络包括多个神经元、特征提取层以及特征映射层，其中，每个神经元与特征映射层相连，所述特征映射层从每个神经元中提取特征信息，所述特征映射层由多个特征向量映射组成。
  5. 如权利要求1所述的基于人工智能的医学影像分类处理系统，其特征在于，所述医学检查影像包括核磁影像、CT影像、超声影像、X光影像以及红外影像。
  6. 一种基于人工智能的医学影像分类处理方法，应用于服务器中，该服务器连接有医学影像信息库、医学影像采集终端和医生工作站，其特征在于，所述基于人工智能的医学影像分类处理方法包括步骤：从医学影像信息库中获取医学影像信息；基于不同人体部位将医学影像信息分类处理；基于分类后的医学影像信息建立对应人体各部位的医学评估模型；对人体各部位的医学评估模型进行训练处理；从训练好的医学评估模型中抽取医学模型参数，并将医学模型参数保存在服务器的存储单元中；从医学影像采集终端接收患者的医学检查影像；从患者的医学检查影像中识别出患者的检查部位，并根据患者的检查部位从存储单元中获取对应的医学模型参数；将患者的医学检查影像及对应的医学模型参数发送到医生工作站供医生对患者进行医学检查。
  7. 如权利要求6所述的基于人工智能的医学影像分类处理方法，其特征在于，所述对人体各部位的医学评估模型进行训练处理的步骤包括：从所述医学评估模型中抽取局部特征信息，并利用人工神经网络对抽取的局部特征信息进行特征训练来获得一组特征向量；将所述特征向量应用于卷积神经网络的训练中，并根据训练结果对医学评估模型进行微调处理。
  8. 如权利要求7所述的基于人工智能的医学影像分类处理方法，其特征在于，所述人工神经网络是通过自动编码器来实现的分层神经网络，该自动编码器将局部特征信息作为第一层的输入并进行特征训练，将第一层的输出作为第二层的输入并进行特征训练，以及将第二层的输出作为第三层的输入并进行特征训练。
  9. 如权利要求7所述的基于人工智能的医学影像分类处理方法，其特征在于，所述卷积神经网络包括多个神经元、特征提取层以及特征映射层，其中，每个神经元与特征映射层相连，所述特征映射层从每个神经元中提取特征信息，所述特征映射层由多个特征向量映射组成。
  10. 如权利要求6所述的基于人工智能的医学影像分类处理方法，其特征在于，所述医学检查影像包括核磁影像、CT影像、超声影像、X光影像以及红外影像。
PCT/CN2017/116664 2017-09-15 2017-12-15 基于人工智能的医学影像分类处理系统及方法 WO2019052063A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710846124.5 2017-09-15
CN201710846124.5A CN107767935A (zh) 2017-09-15 2017-09-15 基于人工智能的医学影像分类处理系统及方法

Publications (1)

Publication Number Publication Date
WO2019052063A1 true WO2019052063A1 (zh) 2019-03-21

Family

ID=61265563

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/116664 WO2019052063A1 (zh) 2017-09-15 2017-12-15 基于人工智能的医学影像分类处理系统及方法

Country Status (2)

Country Link
CN (1) CN107767935A (zh)
WO (1) WO2019052063A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111415727A (zh) * 2020-03-13 2020-07-14 广州医科大学附属肿瘤医院 一种多模态影像数据管理系统及分析方法
CN111739617A (zh) * 2020-06-15 2020-10-02 中国医学科学院医学信息研究所 医疗影像人工智能质控标注方法、装置、设备及存储介质
CN113707289A (zh) * 2021-07-16 2021-11-26 联影智能医疗科技(北京)有限公司 医学人工智能平台及其搭建方法
CN116501904A (zh) * 2023-06-28 2023-07-28 中国人民解放军总医院 分布式存储方法、装置、设备及介质
CN116741350A (zh) * 2023-08-16 2023-09-12 枣庄市山亭区妇幼保健院 一种用于医院x射线影像的档案管理系统

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108922601A (zh) * 2018-07-09 2018-11-30 成都数浪信息科技有限公司 一种医学影像处理系统
CN109346143B (zh) * 2018-12-07 2021-05-11 蔡荣杰 一种提高超声检查报告编辑速度的方法
CN111341402B (zh) * 2018-12-27 2024-04-16 声源康医疗科技(甘肃)有限公司 医学影像信息的存储及交换智慧医疗系统
CN111584074A (zh) * 2019-02-18 2020-08-25 清华大学 肌肉损伤的评价方法及装置
CN110210234A (zh) * 2019-04-23 2019-09-06 平安科技（深圳）有限公司 转诊时医疗信息的迁移方法、装置、计算机设备和存储介质
CN110473607A (zh) * 2019-06-28 2019-11-19 马鞍山师范高等专科学校 一种多终端交互式生物医学影像分析系统
CN115063397A (zh) * 2019-07-18 2022-09-16 上海联影智能医疗科技有限公司 计算机辅助影像分析方法、计算机设备和存储介质
CN112304952B (zh) * 2019-07-23 2024-04-02 纬创资通股份有限公司 影像辨识装置、影像辨识方法及其计算机程序产品
US11798159B2 (en) 2019-09-20 2023-10-24 The Board Of Trustees Of The Leland Stanford Junior University Systems and methods for radiology image classification from noisy images
CN110827958B (zh) * 2019-11-01 2023-04-25 越亮传奇科技股份有限公司 一种医疗影像数据及时采集系统与方法
CN111192679B (zh) * 2019-12-25 2024-04-19 上海联影智能医疗科技有限公司 一种影像数据异常的处理方法、装置及存储介质
CN113053494B (zh) * 2019-12-27 2023-11-21 无锡祥生医疗科技股份有限公司 基于人工智能的pacs系统及其设计方法
CN111582489B (zh) * 2020-05-14 2023-07-14 上海深至信息科技有限公司 一种超声人工智能模型的分布式部署决策系统及方法
CN112686867A (zh) * 2020-12-31 2021-04-20 上海商汤智能科技有限公司 医学图像的识别方法及装置、电子设备和存储介质
CN112951384A (zh) * 2021-02-04 2021-06-11 慧影医疗科技（北京）有限公司 用于医学影像设备的数据模拟发生方法及系统
CN115295125B (zh) * 2022-08-04 2023-11-17 天津市中西医结合医院（天津市南开医院） 一种基于人工智能的医学影像文件管理系统及方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104866727A (zh) * 2015-06-02 2015-08-26 陈宽 基于深度学习对医疗数据进行分析的方法及其智能分析仪
CN106228180A (zh) * 2016-07-14 2016-12-14 杭州健培科技有限公司 一种肺部影像病灶的识别模型参数训练方法及识别方法
CN106355023A (zh) * 2016-08-31 2017-01-25 北京数字精准医疗科技有限公司 基于医学影像的开放式定量分析方法与系统
CN106777953A (zh) * 2016-12-09 2017-05-31 江西中科九峰智慧医疗科技有限公司 医学影像数据的分析方法及系统

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104866727A (zh) * 2015-06-02 2015-08-26 陈宽 基于深度学习对医疗数据进行分析的方法及其智能分析仪
CN106228180A (zh) * 2016-07-14 2016-12-14 杭州健培科技有限公司 一种肺部影像病灶的识别模型参数训练方法及识别方法
CN106355023A (zh) * 2016-08-31 2017-01-25 北京数字精准医疗科技有限公司 基于医学影像的开放式定量分析方法与系统
CN106777953A (zh) * 2016-12-09 2017-05-31 江西中科九峰智慧医疗科技有限公司 医学影像数据的分析方法及系统

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111415727A (zh) * 2020-03-13 2020-07-14 广州医科大学附属肿瘤医院 一种多模态影像数据管理系统及分析方法
CN111739617A (zh) * 2020-06-15 2020-10-02 中国医学科学院医学信息研究所 医疗影像人工智能质控标注方法、装置、设备及存储介质
CN113707289A (zh) * 2021-07-16 2021-11-26 联影智能医疗科技(北京)有限公司 医学人工智能平台及其搭建方法
CN113707289B (zh) * 2021-07-16 2023-11-10 联影智能医疗科技(北京)有限公司 医学人工智能平台及其搭建方法
CN116501904A (zh) * 2023-06-28 2023-07-28 中国人民解放军总医院 分布式存储方法、装置、设备及介质
CN116501904B (zh) * 2023-06-28 2023-09-22 中国人民解放军总医院 分布式存储方法、装置、设备及介质
CN116741350A (zh) * 2023-08-16 2023-09-12 枣庄市山亭区妇幼保健院 一种用于医院x射线影像的档案管理系统
CN116741350B (zh) * 2023-08-16 2023-10-31 枣庄市山亭区妇幼保健院 一种用于医院x射线影像的档案管理系统

Also Published As

Publication number Publication date
CN107767935A (zh) 2018-03-06

Similar Documents

Publication Publication Date Title
WO2019052063A1 (zh) 基于人工智能的医学影像分类处理系统及方法
WO2019052062A1 (zh) 基于人工智能的医学影像报告生成系统及方法
WO2020207377A1 (zh) 一种图像识别模型训练及图像识别方法、装置及系统
CN111310851B (zh) 一种人工智能超声辅助系统及其应用
WO2021093448A1 (zh) 图像处理方法、装置、服务器、医疗图像处理设备及存储介质
CN109993068B (zh) 一种基于心率和面部特征的非接触式的人类情感识别方法
CN110731773B (zh) 一种基于心电波形图全局与局部深度特征融合的异常心电图筛查方法
CN109948719B (zh) 一种基于残差密集模块网络结构的眼底图像质量自动分类方法
CN110584618B (zh) 基于人工智能的传染病机器识别系统
CN110164550B (zh) 一种基于多视角协同关系的先天性心脏病辅助诊断方法
CN114188021B (zh) 一种基于多模态融合的儿童肠套叠诊断智能分析系统
Li et al. PSENet: Psoriasis severity evaluation network
CN113116300A (zh) 一种基于模型融合的生理信号分类方法
Zhang et al. A survey of wound image analysis using deep learning: Classification, detection, and segmentation
CN117542474A (zh) 基于大数据的远程护理监测系统及方法
CN113397485A (zh) 一种基于深度学习的脊柱侧弯筛查方法
CN113506274B (zh) 基于视觉显著性差异图的用于人体认知状况的检测系统
CN113298773A (zh) 基于深度学习的心脏视图识别与左心室检测装置、系统
Jeon et al. Multi-slice Nested Recurrence Plot (MsNRP): A robust approach for person identification using daily ECG or PPG signals
CN116978510A (zh) 一种居家脑卒中患者的护理分级评估系统
CN116664592A (zh) 基于图像的动静脉血管分离方法、装置、电子设备及介质
CN116189884A (zh) 一种基于面部视觉的多模态融合中医体质判别方法及系统
Kalbande et al. An artificial intelligence approach for the recognition of early stages of eczema
CN113796850A (zh) 甲状旁腺mibi图像分析系统、计算机设备及存储介质
Qi et al. Residual neural networks based on empirical mode decomposition for mitral regurgitation prediction

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: 17925513

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 17/09/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 17925513

Country of ref document: EP

Kind code of ref document: A1