WO2021115130A1 - 脑出血点智能检测方法、装置、电子设备及存储介质 - Google Patents

脑出血点智能检测方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2021115130A1
WO2021115130A1 PCT/CN2020/131904 CN2020131904W WO2021115130A1 WO 2021115130 A1 WO2021115130 A1 WO 2021115130A1 CN 2020131904 W CN2020131904 W CN 2020131904W WO 2021115130 A1 WO2021115130 A1 WO 2021115130A1
Authority
WO
WIPO (PCT)
Prior art keywords
brain
hemorrhage
detection model
image
original
Prior art date
Application number
PCT/CN2020/131904
Other languages
English (en)
French (fr)
Inventor
杨光
Original Assignee
平安科技（深圳）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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021115130A1 publication Critical patent/WO2021115130A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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; CALCULATING OR 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]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30016—Brain
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device and computer-readable storage medium for intelligent detection of cerebral hemorrhage points.
  • the operating platform for brain hemorrhage detection in the market is not perfect, and most of them are based on traditional image detection algorithms.
  • the intelligent detection method of cerebral hemorrhage point includes:
  • Receive the brain picture input by the user input the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, and edit and organize the brain hemorrhage point coordinates based on the pre-built hemorrhage point operation platform and the brain hemorrhage point coordinates Output brain bleeding points.
  • the present application also provides an intelligent detection device for cerebral hemorrhage point, the device includes:
  • Data receiving and processing module used to obtain brain medical image sets, and perform purification operations and format conversions on the brain medical image sets based on the nature of cerebral hemorrhage to generate original readable brain image sets;
  • Image adjustment module used to adjust the original readable brain image set based on a pre-built parameter adjustment algorithm to obtain a readable brain image set;
  • Bleeding point detection training model module used to divide the readable brain image set into a training set and a test set, and input the training set into the pre-built original brain hemorrhage point detection model to train to obtain brain hemorrhage points Detection model, input the test set to the brain hemorrhage point detection model to obtain a test value, if the test value is less than a preset threshold, the brain hemorrhage point detection model receives the training set to continue training, if all The test value is greater than a preset threshold, and the brain hemorrhage point detection model completes the training;
  • Hemorrhage point result output module receives the brain picture input by the user, inputs the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, based on the pre-built hemorrhage point operation platform and the brain Edit the bleeding point coordinates and output the brain bleeding point.
  • the present application also provides an electronic device, the electronic device including a memory and a processor, the memory stores a brain hemorrhage point intelligent detection program that can be run on the processor, and the brain hemorrhage point intelligent detection program is When the processor executes, the following steps are implemented:
  • Receive the brain picture input by the user input the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, and edit and organize the brain hemorrhage point coordinates based on the pre-built hemorrhage point operation platform and the brain hemorrhage point coordinates Output brain bleeding points.
  • the present application also provides a computer-readable storage medium on which a brain hemorrhage point intelligent detection program is stored, and the brain hemorrhage point intelligent detection program can be executed by one or more processors to achieve the following step:
  • Receive the brain picture input by the user input the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, and edit and organize the brain hemorrhage point coordinates based on the pre-built hemorrhage point operation platform and the brain hemorrhage point coordinates Output brain bleeding points.
  • FIG. 1 is a schematic flowchart of a method for intelligent detection of cerebral hemorrhage points according to an embodiment of the application
  • FIG. 2 is a schematic diagram of the internal structure of an electronic device provided by an embodiment of the application.
  • FIG. 3 is a schematic diagram of modules of an intelligent detection device for cerebral hemorrhage points provided by an embodiment of the application.
  • FIG. 1 it is a schematic flowchart of a method for intelligent detection of cerebral hemorrhage points according to an embodiment of this application.
  • the method can be executed by a device, and the device can be implemented by software and/or hardware.
  • the intelligent detection method of cerebral hemorrhage point includes:
  • the preferred embodiment of the present application receives the brain medical image collection from the PACS (Picture Archiving and Communication Systems) of the hospital. Furthermore, this application can also accept manual case screening, sorting out and screening brain medical images that have undergone brain CT examinations in the hospital due to cerebral hemorrhage.
  • PACS Picture Archiving and Communication Systems
  • the purification operation includes: automatically calibrating the brain tissue area, cerebrospinal fluid, ventricle, and lateral skull in the brain medical image concentration, and extracting the brain tissue area in the brain medical image concentration as light gray and the cerebrospinal fluid And a collection of pictures in which the cerebral ventricle is black and the lateral skull is white.
  • the brain medical image set uses the number of bits to store the designated image storage value
  • the computer image file is composed of three primary colors of red, green, and blue, and the number of levels that can be displayed as a grayscale image is 0- 255, so it is necessary to convert the number of bits of the specified image storage value into a three-primary color image.
  • the conversion formula is:
  • v is the image storage value of the brain medical image collection
  • G is the display value on the computer operating platform
  • gm is the maximum bitmap file displayed on the computer monitor is 255
  • w is the window of the brain medical image collection Wide
  • c is the window level of the brain medical image collection.
  • the window width is the CT/DR value of the image display, and the window level is the center coordinate position of the window.
  • the parameter adjustment algorithm includes a vector generation function, an image extraction function, an image display function, and a gray scale stretch function.
  • adjusting the original readable brain image set based on a pre-built parameter adjustment algorithm to obtain the window width and window position of the readable brain image set includes: converting the original readable brain image set based on the vector generation function
  • the pixel data of the image set is stored in an unsigned integer array, and the unsigned integer array is set as a vector set according to a preset scale, and cerebral hemorrhage image data is extracted from the vector set according to the image extraction function, and Image display is performed based on the image display function, and according to the image display, the gray-scale stretching function is adjusted to obtain an adjusted readable brain image set.
  • the vector generation function may use the getUint16Array() function that comes with the programming language
  • the image extraction function image and the gray scale stretch function may be based on the DCMTK library developed in Germany
  • the display function may be based on the QT library .
  • inputting the training set into a pre-built original brain hemorrhage point detection model to train to obtain a brain hemorrhage point detection model includes: marking the brain hemorrhage point out of the training set to obtain a real hemorrhage point label Coordinate set, initialize the model parameters of the original brain hemorrhage point detection model, input the training set into the original brain hemorrhage point detection model, and the original brain hemorrhage point detection model continuously according to the training set Adjust the model parameters and output the predicted bleeding point mark coordinate set, and determine whether the error between the predicted bleeding point mark coordinate set and the real bleeding point mark coordinate set is less than a preset error. If it is greater than the preset error, the The original brain hemorrhage point detection model continues to adjust the model parameters, and if it is less than the preset error, the training is completed to obtain the brain hemorrhage point detection model.
  • the original brain bleeding spot detection model is based on a convolutional neural network model, and the model parameters include contrast, brightness, rolling adjacent brain image slices, and confidence parameters of the bleeding spot area.
  • the convolutional neural network model includes a convolution operation and a pooling operation. Further, the convolution operation and the pooling operation include pre-building a convolution template and determining a convolution step size, and according to the convolution step size, The convolution template is calculated with the training set to obtain a convolution matrix set after the convolution operation, and the convolution operation is completed. The maximum value or the average value of the matrix in the convolution matrix set is selected to replace the convolution matrix set to complete the pooling operation.
  • the pre-built convolution template may be a standard 3*3 matrix, such as The calculation method of the matrix obtained by the calculation of the convolution operation is from left to right, and the convolution amplitude is 1.
  • the feature candidate region matrix with a feature of 9*9 in the feature candidate region is as follows: The pre-built convolution template First with The calculation method is: 1*0, 0*3, 1*1 and other corresponding dimensions are multiplied, and the final result is:
  • the pre-built convolution template According to the convolution amplitude of 1, continue to move one step to the right and the matrix is: the pre-built convolution template Perform the above operations to obtain the pre-built convolution template It can be seen that when the convolution operation is completed, a large number of small-dimensional matrices can be generated, as described above with Therefore, the pooling operation is to reduce the dimensions of a large number of small-dimensional matrices generated by the convolution operation.
  • the principle of maximization can be adopted, as described above with Use the largest
  • the above-mentioned convolution operation and pooling operation are repeated, preferably 16 times of the convolution and pooling operation can be used to obtain the predicted bleeding point mark coordinate set.
  • the bleeding point operation platform can display the detection results of multiple brain bleeding point detection models in different periods in real time and synchronously. Based on the parallel display, the user can not only display multiple different detection result versions It also provides a variety of editing methods to directly modify or adjust the detected contours.
  • the adjusted contour data can continue to be used as the training set of step S3 and input into the brain hemorrhage detection model for training to speed up. The efficiency of verification, retraining, and reverification of the detection result of the cerebral hemorrhage point detection model is improved, and the efficiency is improved.
  • the invention also provides an electronic device.
  • FIG. 2 it is a schematic diagram of the internal structure of an electronic device provided by an embodiment of this application.
  • the electronic device 1 may be a PC (Personal Computer, personal computer), or a terminal device such as a smart phone, a tablet computer, or a portable computer, or a server.
  • the electronic device 1 at least includes a memory 11, a processor 12, a communication bus 13, and a network interface 14.
  • the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, and the like.
  • the memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a hard disk of the electronic device 1.
  • the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in hard disk equipped on the electronic device 1, a smart media card (SMC), or a secure digital (SD) Card, Flash Card, etc.
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can be used not only to store application software and various types of data installed in the electronic device 1, such as codes for a brain hemorrhage point intelligent detection program, etc., but also to temporarily store data that has been output or will be output.
  • the processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip, for running program codes or processing stored in the memory 11 Data, such as the implementation of intelligent detection procedures for cerebral hemorrhage points.
  • CPU central processing unit
  • controller microcontroller
  • microprocessor microprocessor
  • other data processing chip for running program codes or processing stored in the memory 11 Data, such as the implementation of intelligent detection procedures for cerebral hemorrhage points.
  • the communication bus 13 is used to realize the connection and communication between these components.
  • the network interface 14 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface), and is usually used to establish a communication connection between the device 1 and other electronic devices.
  • the device 1 may also include a user interface.
  • the user interface may include a display (Display) and an input unit such as a keyboard (Keyboard).
  • the optional user interface may also include a standard wired interface and a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc.
  • the display can also be appropriately called a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visualized user interface.
  • Figure 2 only shows the electronic device 1 with components 11-14 and a brain hemorrhage point intelligent detection program.
  • the structure shown in Figure 1 does not constitute a limitation on the electronic device 1, and may include Fewer or more parts than shown, or some parts in combination, or different parts arrangement.
  • the memory 11 stores a brain hemorrhage point intelligent detection program; when the processor 12 executes the brain hemorrhage point intelligent detection program stored in the memory 11, the following steps are implemented:
  • Step 1 Obtain a brain medical image set, and perform a purification operation and format conversion on the brain medical image set based on the nature of cerebral hemorrhage to generate an original readable brain image set.
  • the preferred embodiment of the present application receives the brain medical image collection from the PACS (Picture Archiving and Communication Systems) of the hospital. Furthermore, this application can also accept manual case screening, sorting out and screening brain medical images that have undergone brain CT examinations in the hospital due to cerebral hemorrhage.
  • PACS Picture Archiving and Communication Systems
  • the purification operation includes: automatically calibrating the brain tissue area, cerebrospinal fluid, ventricle, and lateral skull in the brain medical image concentration, and extracting the brain tissue area in the brain medical image concentration as light gray and the cerebrospinal fluid And a collection of pictures in which the cerebral ventricle is black and the lateral skull is white.
  • the brain medical image set uses the number of bits to store the designated image storage value
  • the computer image file is composed of three primary colors of red, green, and blue, and the number of levels that can be displayed as a grayscale image is 0- 255, so it is necessary to convert the number of bits of the specified image storage value into a three-primary color image.
  • the conversion formula is:
  • v is the image storage value of the brain medical image collection
  • G is the display value on the computer operating platform
  • gm is the maximum bitmap file displayed on the computer monitor is 255
  • w is the window of the brain medical image collection Wide
  • c is the window level of the brain medical image collection.
  • the window width is the CT/DR value of the image display, and the window level is the center coordinate position of the window.
  • Step 2 Adjust the original readable brain image set based on the pre-built parameter adjustment algorithm to obtain a readable brain image set.
  • the parameter adjustment algorithm includes a vector generation function, an image extraction function, an image display function, and a gray scale stretching function.
  • adjusting the original readable brain image set based on a pre-built parameter adjustment algorithm to obtain the window width and window position of the readable brain image set includes: converting the original readable brain image set based on the vector generation function
  • the pixel data of the image set is stored in an unsigned integer array, and the unsigned integer array is set as a vector set according to a preset scale, and cerebral hemorrhage image data is extracted from the vector set according to the image extraction function, and Image display is performed based on the image display function, and according to the image display, the gray-scale stretching function is adjusted to obtain an adjusted readable brain image set.
  • the vector generation function may use the getUint16Array() function that comes with the programming language
  • the image extraction function image and the gray scale stretch function may be based on the DCMTK library developed in Germany
  • the display function may be based on the QT library .
  • Step 3 Divide the readable brain image set into a training set and a test set, and input the training set into the pre-built original brain hemorrhage detection model to train to obtain the brain hemorrhage detection model.
  • the test set is input to the brain bleeding point detection model to obtain a test value. If the test value is less than a preset threshold, the brain bleeding point detection model receives the training set and continues training; if the test value is greater than the preset threshold Threshold, the brain hemorrhage point detection model completes the training.
  • inputting the training set into a pre-built original brain hemorrhage point detection model to train to obtain a brain hemorrhage point detection model includes: marking the brain hemorrhage point out of the training set to obtain a real hemorrhage point label Coordinate set, initialize the model parameters of the original brain hemorrhage point detection model, input the training set into the original brain hemorrhage point detection model, and the original brain hemorrhage point detection model continuously according to the training set Adjust the model parameters and output the predicted bleeding point mark coordinate set, and determine whether the error between the predicted bleeding point mark coordinate set and the real bleeding point mark coordinate set is less than a preset error. If it is greater than the preset error, the The original brain hemorrhage point detection model continues to adjust the model parameters, and if it is less than the preset error, the training is completed to obtain the brain hemorrhage point detection model.
  • the original brain bleeding spot detection model is based on a convolutional neural network model, and the model parameters include contrast, brightness, rolling adjacent brain image slices, and confidence parameters of the bleeding spot area.
  • the convolutional neural network model includes a convolution operation and a pooling operation. Further, the convolution operation and the pooling operation include pre-building a convolution template and determining a convolution step size, and according to the convolution step size, The convolution template is calculated with the training set to obtain a convolution matrix set after the convolution operation, and the convolution operation is completed. The maximum value or the average value of the matrix in the convolution matrix set is selected to replace the convolution matrix set to complete the pooling operation.
  • the pre-built convolution template may be a standard 3*3 matrix, such as The calculation method of the matrix obtained by the calculation of the convolution operation is from left to right, and the convolution amplitude is 1.
  • the feature candidate region matrix with a feature of 9*9 in the feature candidate region is as follows: The pre-built convolution template First with The calculation method is: 1*0, 0*3, 1*1 and other corresponding dimensions are multiplied, and the final result is:
  • the pre-built convolution template According to the convolution amplitude of 1, continue to move one step to the right and the matrix is: the pre-built convolution template Perform the above operations to obtain the pre-built convolution template It can be seen that when the convolution operation is completed, a large number of small-dimensional matrices can be generated, as described above with Therefore, the pooling operation is to reduce the dimensions of a large number of small-dimensional matrices generated by the convolution operation.
  • the principle of maximization can be adopted, as described above with Use the largest
  • the above-mentioned convolution operation and pooling operation are repeated, preferably 16 times of the convolution and pooling operation can be used to obtain the predicted bleeding point mark coordinate set.
  • Step 4 Receive the brain picture input by the user, input the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, based on the pre-built hemorrhage point operation platform and the brain hemorrhage point coordinates Edit and sort out brain bleeding points.
  • the bleeding point operation platform can display the detection results of multiple brain bleeding point detection models in different periods in real time and synchronously. Based on the parallel display, the user can not only display multiple different detection result versions It also provides a variety of editing methods to directly modify or adjust the detected contours.
  • the adjusted contour data can continue to be used as the training set of step 3 and input into the brain bleeding point detection model for training to speed up
  • the efficiency of verification, retraining, and reverification of the detection result of the cerebral hemorrhage point detection model is improved, and the efficiency is improved.
  • FIG. 3 is a schematic diagram of modules in an embodiment of an intelligent detection device for cerebral hemorrhage points of this application
  • the intelligent detection device 2 for cerebral hemorrhage points can be divided into data receiving and processing modules 10,
  • the image adjustment module 20, the bleeding point detection training model module 30, and the bleeding point result output module 40 are exemplary:
  • the data receiving and processing module 10 is used to obtain a brain medical image set, and perform a purification operation and format conversion on the brain medical image set based on the nature of cerebral hemorrhage to generate an original readable brain image set.
  • the image adjustment module 20 is configured to adjust the original readable brain image set based on a pre-built parameter adjustment algorithm to obtain a readable brain image set.
  • the bleeding point detection training model module 30 is used to: divide the readable brain image set into a training set and a test set, and input the training set into a pre-built original brain bleeding point detection model to train the brain.
  • a bleeding point detection model the test set is input to the brain bleeding point detection model to obtain a test value, and if the test value is less than a preset threshold, the brain bleeding point detection model receives the training set and continues training If the test value is greater than the preset threshold, the brain bleeding point detection model completes the training.
  • the bleeding point result output module 40 is configured to: receive a brain picture input by a user, and input the brain picture into the brain bleeding point detection model to obtain the brain bleeding point coordinates, based on a pre-built bleeding point operating platform Edit and sort the coordinates of the cerebral hemorrhage point to output the cerebral hemorrhage point.
  • the embodiment of the present application also proposes a computer-readable storage medium.
  • the computer-readable storage medium may be non-volatile or volatile.
  • the computer-readable storage medium stores brain hemorrhage point intelligence.
  • the detection program, the brain hemorrhage point intelligent detection program can be executed by one or more processors to achieve the following operations:
  • the original readable brain image set is adjusted based on a pre-built parameter adjustment algorithm to obtain a readable brain image set.
  • the brain hemorrhage point detection model Divides the readable brain image set into a training set and a test set, input the training set into a pre-built original brain hemorrhage detection model to train a brain hemorrhage detection model, and input the test set Until the brain hemorrhage point detection model obtains a test value, if the test value is less than the preset threshold, the brain hemorrhage point detection model receives the training set to continue training, and if the test value is greater than the preset threshold, so The brain bleeding point detection model completes the training.
  • Receive the brain picture input by the user input the brain picture into the brain hemorrhage point detection model to obtain the brain hemorrhage point coordinates, and edit and organize the brain hemorrhage point coordinates based on the pre-built hemorrhage point operation platform and the brain hemorrhage point coordinates Output brain bleeding points.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Public Health (AREA)
  • Image Processing (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

一种人工智能技术，揭露了一种脑出血点智能检测方法，包括：获取脑部医疗影像集并基于参数调整算法得到可读脑部图像集，将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，直至所述测试值满足预设要求，所述脑部出血点检测模型完成所述训练，接收脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。还包括一种脑出血点智能检测装置、电子设备以及一种计算机可读存储介质。实现精准高效的脑出血点检测功能。

Description

脑出血点智能检测方法、装置、电子设备及存储介质
本申请要求于2019年12月9日提交中国专利局、申请号为201911254426.9，发明名称为“脑出血点智能检测方法、装置及计算机可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，尤其涉及一种脑出血点智能检测的方法、装置及计算机可读存储介质。
背景技术
目前市面上对于脑部出血点检测的操作平台不够完善，且多数基于传统图像检测算法，发明人意识到传统图像检测算法具有检测效率低，检测准确率低等原因，故而不能很好辅助医生快速找到出血点，导致患者因出血点不能快速定位而延误治疗。
发明内容
本申请提供的一种脑出血点智能检测方法，包括：
获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
本申请还提供一种脑出血点智能检测装置，所述装置包括：
数据接收及处理模块：用于获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
图像调整模块：用于基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
出血点检测训练模型模块：用于将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
出血点结果输出模块：接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
本申请还提供一种电子设备，所述电子设备包括存储器和处理器，所述存储 器中存储有可在所述处理器上运行的脑出血点智能检测程序，所述脑出血点智能检测程序被所述处理器执行时实现如下步骤：
获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
本申请还提供一种计算机可读存储介质，所述计算机可读存储介质上存储有脑出血点智能检测程序，所述脑出血点智能检测程序可被一个或者多个处理器执行，以实现如下步骤：
获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
附图说明
图1为本申请一实施例提供的脑出血点智能检测方法的流程示意图；
图2为本申请一实施例提供的电子设备的内部结构示意图；
图3为本申请一实施例提供的脑出血点智能检测装置的模块示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请提供一种脑出血点智能检测方法。参照图1所示，为本申请一实施例提供的脑出血点智能检测方法的流程示意图。该方法可以由一个装置执行，该装置可以由软件和/或硬件实现。
在本实施例中，脑出血点智能检测方法包括：
S1、获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集。
本申请较佳实施例从医院的PACS(Picture Archiving and Communication Systems，影像归档和通信系统)接收所述脑部医疗影像集。更进一步，本申请也可接收通过人工进行病例筛选，整理并筛选因脑出血而在医院接受脑CT检查的脑部医疗影像。
优选地，所述提纯操作包括：自动化标定所述脑部医疗影像集中脑组织区域、脑脊液、脑室及外侧颅骨，提取出所述脑部医疗影像集中所述脑组织区域为浅灰色、所述脑脊液及所述脑室为黑色、所述外侧颅骨为白色的图片集。
进一步地，所述脑部医疗影像集采用位存储指定图像存储值的位数，而计算机的图像文件是由红、绿、蓝三基色组成，作为灰度图像所能显示的等级数在0-255之间，因此需要将所述位存储指定图像存储值的位数转换为三基色图像。优选地，转换公式为：
Figure PCTCN2020131904-appb-000001
其中，v为所述脑部医疗影像集的图像存储值，G为计算机操作平台上的显示值，gm为计算机显示器显示的最大位图文件为255，w为所述脑部医疗影像集的窗宽，c为所述脑部医疗影像集的窗位。
所述窗宽是图像显示的CT/DR值，所述窗位是窗的中心坐标位置。
S2、基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集。
优选地，所述参数调整算法包括向量生成函数、图像提取函数、图像显示函数和灰度拉伸函数。
较佳地，基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集的窗宽窗位，包括：基于所述向量生成函数，将所述原始可读脑部图像集的像素数据存储到无符号整型数组，并将所述无符号整型数组按预设规模设置为向量集，根据所述图像提取函数从所述向量集中提取脑出血图像数据，并基于所述图像显示函数进行图像显示，根据所述图像显示，所述灰度拉伸函数进行参数调节得到调节后的可读脑部图像集。
进一步地，所述向量生成函数可采用编程语言自带的getUint16Array()函数，所述图像提取函数图像和所述灰度拉伸函数可基于德国开发的DCMTK库，所述显示函数可基于QT库。
S3、将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练。
优选地，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，包括：将所述训练集进行出脑部出血点标记，得到真实出血点标记坐标集，初始化所述原始脑部出血点检测模型的模型参数，将所述训练集输入至所述原始脑部出血点检测模型中，所述原始脑部出血点检测模型根据 所述训练集不断调整所述模型参数并输出预测出血点标记坐标集，判断所述预测出血点标记坐标集与所述真实出血点标记坐标集的误差是否小于预设误差，若大于所述预设误差，所述原始脑部出血点检测模型继续调整所述模型参数，若小于所述预设误差，完成所述训练得到所述脑部出血点检测模型。
所述原始脑部出血点检测模型是基于卷积神经网络模型，所述模型参数包括对比度、亮度、滚动相邻的脑部图像切片，出血点区域的置信度参数等。
所述卷积神经网络模型包括卷积操作和池化操作，进一步地，所述卷积操作和池化操作包括预先构建卷积模板并确定卷积步长，根据所述卷积步长，将卷积模板与所述训练集进行计算得到卷积操作后的卷积矩阵集，完成所述卷积操作。选择所述卷积矩阵集中矩阵的最大值或平均值替代所述卷积矩阵集，完成所述池化操作。
进一步地，所述预先构建卷积模板可为标准的3*3矩阵，如
Figure PCTCN2020131904-appb-000002
所述计算得到卷积操作后的矩阵的计算方式是采用从左至右，卷积幅度为1的方式，如所述特征候选区域集中有特征为9*9的特征候选区域矩阵为：
Figure PCTCN2020131904-appb-000003
则所述预先构建卷积模板
Figure PCTCN2020131904-appb-000004
先与
Figure PCTCN2020131904-appb-000005
进行计算，计算方式为：1*0、0*3、1*1等对应维度相乘，最终得到的结果为：
Figure PCTCN2020131904-appb-000006
依次类推，所述预先构建卷积模板
Figure PCTCN2020131904-appb-000007
根据卷积幅度为1，向右继续横移一步与矩阵为：所述预先构建卷积模板
Figure PCTCN2020131904-appb-000008
进行上述操作得到所述预先构建卷积模板
Figure PCTCN2020131904-appb-000009
由此可知，当完成所述卷积操作可生成大量的小维度矩阵，如上述
Figure PCTCN2020131904-appb-000010
和
Figure PCTCN2020131904-appb-000011
等，因此，所述池化操作是将所述卷积操作生成的大量的小维度矩阵的维度变小，较佳地可采用最大化原理，如将上述
Figure PCTCN2020131904-appb-000012
和
Figure PCTCN2020131904-appb-000013
用最大的数值3和7代替，从而完成所述池化操作。
优选地，反复进行上述卷积操作和池化操作，较佳地可使用16次所述卷积和池化操作后得到所述预测出血点标记坐标集。
S4、接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
较佳地，所述出血点操作平台可以实时并同步显示多个所述脑部出血点检测模型在不同时期的检测结果，在并行显示的基础上，用户不但可以在多个不同的检测结果版本上进行对比，而且提供了多种编辑方式可直接修改或者调整检测出的轮廓，调整后的轮廓数据可以继续作为S3步骤的训练集并输入至所述脑部出血点检测模型中进行训练，加快了所述脑部出血点检测模型检测结果的验证、再训练、再验证的效率，提高了效率。
发明还提供一种电子设备。参照图2所示，为本申请一实施例提供的电子设备的内部结构示意图。
在本实施例中，所述电子设备1可以是PC(Personal Computer，个人电脑)，或者是智能手机、平板电脑、便携计算机等终端设备，也可以是一种服务器等。该电子设备1至少包括存储器11、处理器12，通信总线13，以及网络接口14。
其中，存储器11至少包括一种类型的可读存储介质，所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如，SD或DX存储器等)、磁性存储器、磁盘、光盘等。存储器11在一些实施例中可以是电子设备1的内部存储单元，例如该电子设备1的硬盘。存储器11在另一些实施例中也可以是电子设备1的外部存储设备，例如电子设备1上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。进一步地，存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据，例如脑出血点智能检测程序的代码等，还可以用于暂时地存储已经输出或者将要输出的数据。
处理器12在一些实施例中可以是一中央处理器(Central Processing Unit，CPU)、控制器、微控制器、微处理器或其他数据处理芯片，用于运行存储器11中存储的程序代码或处理数据，例如执行脑出血点智能检测程序等。
通信总线13用于实现这些组件之间的连接通信。
网络接口14可选的可以包括标准的有线接口、无线接口(如WI-FI接口)，通常用于在该装置1与其他电子设备之间建立通信连接。
可选地，该装置1还可以包括用户接口，用户接口可以包括显示器(Display)、输入单元比如键盘(Keyboard)，可选的用户接口还可以包括标准的有线接口、无线接口。可选地，在一些实施例中，显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode，有机发光二极管)触摸器等。其中，显示器也可以适当的称为显示屏或显示单元，用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
图2仅示出了具有组件11-14以及脑出血点智能检测程序的电子设备1，本领域技术人员可以理解的是，图1示出的结构并不构成对电子设备1的限定，可以包括比图示更少或者更多的部件，或者组合某些部件，或者不同的部件布置。
在图2所示的装置1实施例中，存储器11中存储有脑出血点智能检测程序；处理器12执行存储器11中存储的脑出血点智能检测程序时实现如下步骤：
步骤一、获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集。
本申请较佳实施例从医院的PACS(Picture Archiving and Communication  Systems，影像归档和通信系统)接收所述脑部医疗影像集。更进一步，本申请也可接收通过人工进行病例筛选，整理并筛选因脑出血而在医院接受脑CT检查的脑部医疗影像。
优选地，所述提纯操作包括：自动化标定所述脑部医疗影像集中脑组织区域、脑脊液、脑室及外侧颅骨，提取出所述脑部医疗影像集中所述脑组织区域为浅灰色、所述脑脊液及所述脑室为黑色、所述外侧颅骨为白色的图片集。
进一步地，所述脑部医疗影像集采用位存储指定图像存储值的位数，而计算机的图像文件是由红、绿、蓝三基色组成，作为灰度图像所能显示的等级数在0-255之间，因此需要将所述位存储指定图像存储值的位数转换为三基色图像。优选地，转换公式为：
Figure PCTCN2020131904-appb-000014
其中，v为所述脑部医疗影像集的图像存储值，G为计算机操作平台上的显示值，gm为计算机显示器显示的最大位图文件为255，w为所述脑部医疗影像集的窗宽，c为所述脑部医疗影像集的窗位。
所述窗宽是图像显示的CT/DR值，所述窗位是窗的中心坐标位置。
步骤二、基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集。
优选地，所述参数调整算法包括向量生成函数、图像提取函数、图像显示函数和灰度拉伸函数。
较佳地，基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集的窗宽窗位，包括：基于所述向量生成函数，将所述原始可读脑部图像集的像素数据存储到无符号整型数组，并将所述无符号整型数组按预设规模设置为向量集，根据所述图像提取函数从所述向量集中提取脑出血图像数据，并基于所述图像显示函数进行图像显示，根据所述图像显示，所述灰度拉伸函数进行参数调节得到调节后的可读脑部图像集。
进一步地，所述向量生成函数可采用编程语言自带的getUint16Array()函数，所述图像提取函数图像和所述灰度拉伸函数可基于德国开发的DCMTK库，所述显示函数可基于QT库。
步骤三、将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练。
优选地，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，包括：将所述训练集进行出脑部出血点标记，得到真实出血点标记坐标集，初始化所述原始脑部出血点检测模型的模型参数，将所述训练集输入至所述原始脑部出血点检测模型中，所述原始脑部出血点检测模型根据所述训练集不断调整所述模型参数并输出预测出血点标记坐标集，判断所述预测出血点标记坐标集与所述真实出血点标记坐标集的误差是否小于预设误差，若大 于所述预设误差，所述原始脑部出血点检测模型继续调整所述模型参数，若小于所述预设误差，完成所述训练得到所述脑部出血点检测模型。
所述原始脑部出血点检测模型是基于卷积神经网络模型，所述模型参数包括对比度、亮度、滚动相邻的脑部图像切片，出血点区域的置信度参数等。
所述卷积神经网络模型包括卷积操作和池化操作，进一步地，所述卷积操作和池化操作包括预先构建卷积模板并确定卷积步长，根据所述卷积步长，将卷积模板与所述训练集进行计算得到卷积操作后的卷积矩阵集，完成所述卷积操作。选择所述卷积矩阵集中矩阵的最大值或平均值替代所述卷积矩阵集，完成所述池化操作。
进一步地，所述预先构建卷积模板可为标准的3*3矩阵，如
Figure PCTCN2020131904-appb-000015
所述计算得到卷积操作后的矩阵的计算方式是采用从左至右，卷积幅度为1的方式，如所述特征候选区域集中有特征为9*9的特征候选区域矩阵为：
Figure PCTCN2020131904-appb-000016
则所述预先构建卷积模板
Figure PCTCN2020131904-appb-000017
先与
Figure PCTCN2020131904-appb-000018
进行计算，计算方式为：1*0、0*3、1*1等对应维度相乘，最终得到的结果为：
Figure PCTCN2020131904-appb-000019
依次类推，所述预先构建卷积模板
Figure PCTCN2020131904-appb-000020
根据卷积幅度为1，向右继续横移一步与矩阵为：所述预先构建卷积模板
Figure PCTCN2020131904-appb-000021
进行上述操作得到所述预先构建卷积模板
Figure PCTCN2020131904-appb-000022
由此可知，当完成所述卷积操作可生成大量的小维度矩阵，如上述
Figure PCTCN2020131904-appb-000023
和
Figure PCTCN2020131904-appb-000024
等，因此，所述池化操作是将所述卷积操作生成的大量的小维度矩阵的维度变小，较佳地可采用最大化原理，如将上述
Figure PCTCN2020131904-appb-000025
和
Figure PCTCN2020131904-appb-000026
用最大的数值3和7代替，从而完成所述池化操作。
优选地，反复进行上述卷积操作和池化操作，较佳地可使用16次所述卷积和池化操作后得到所述预测出血点标记坐标集。
步骤四、接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血 点坐标进行编辑整理输出脑部出血点。
较佳地，所述出血点操作平台可以实时并同步显示多个所述脑部出血点检测模型在不同时期的检测结果，在并行显示的基础上，用户不但可以在多个不同的检测结果版本上进行对比，而且提供了多种编辑方式可直接修改或者调整检测出的轮廓，调整后的轮廓数据可以继续作为步骤三的训练集并输入至所述脑部出血点检测模型中进行训练，加快了所述脑部出血点检测模型检测结果的验证、再训练、再验证的效率，提高了效率。
例如，参照图3所示，为本申请脑出血点智能检测装置一实施例中的模块示意图，该实施例中，所述脑出血点智能检测装置2可以被分割为数据接收及处理模块10、图像调整模块20、出血点检测训练模型模块30、出血点结果输出模块40示例性地：
所述数据接收及处理模块10用于：获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集。
所述图像调整模块20用于：基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集。
所述出血点检测训练模型模块30用于：将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练。
所述出血点结果输出模块40用于：接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
上述数据接收及处理模块10、图像调整模块20、出血点检测训练模型模块30、出血点结果输出模块40等程序模块被执行时所实现的功能或操作步骤与上述实施例大体相同，在此不再赘述。
此外，本申请实施例还提出一种计算机可读存储介质，所述计算机可读存储介质可以是非易失性，也可以是易失性，所述计算机可读存储介质上存储有脑出血点智能检测程序，所述脑出血点智能检测程序可被一个或多个处理器执行，以实现如下操作：
获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集。
基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集。
将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练。
接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
需要说明的是，上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。并且本文中的术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含， 从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、装置、物品或者方法所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，或者网络设备等)执行本申请各个实施例所述的方法。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种脑出血点智能检测方法，其中，所述方法包括：
    获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
    基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
    将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
    接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
  2. 如权利要求1所述的脑出血点智能检测方法，其中，所述格式转换的计算方法为：
    Figure PCTCN2020131904-appb-100001
    其中，v为所述脑部医疗影像集的图像存储值，G为计算机操作平台上的显示值，gm为计算机显示器显示的最大位图文件为255，w为所述脑部医疗影像集的窗宽，c为所述脑部医疗影像集的窗位。
  3. 如权利要求1所述的脑出血点智能检测方法，其中，所述参数调整算法包括向量生成函数、图像提取函数、图像显示函数和灰度拉伸函数；
    所述基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集，包括：
    基于所述向量生成函数，将所述原始可读脑部图像集的像素数据存储到无符号整型数组，并将所述无符号整型数组转变为向量集；
    根据所述图像提取函数从所述向量集中提取脑出血图像数据，并基于所述图像显示函数进行图像显示；
    根据所述图像显示结果，调用所述灰度拉伸函数进行参数调节得到调节后的可读脑部图像集。
  4. 如权利要求1所述的脑出血点智能检测方法，其中，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，包括：
    将所述训练集进行脑部出血点标记，得到真实出血点标记坐标集；
    初始化所述原始脑部出血点检测模型的模型参数；
    将所述训练集输入至所述原始脑部出血点检测模型中，所述原始脑部出血点检测模型根据所述训练集不断调整所述模型参数并输出预测出血点标记坐标集；
    判断所述预测出血点标记坐标集与所述真实出血点标记坐标集的误差是否小于预设误差，若大于所述预设误差，所述原始脑部出血点检测模型继续调整所 述模型参数，若小于所述预设误差，完成所述训练得到所述脑部出血点检测模型。
  5. 如权利要求1至4中任意一项所述的脑出血点智能检测方法，其中，所述提纯操作包括：
    自动化标定所述脑部医疗影像集中脑组织区域、脑脊液、脑室及外侧颅骨；
    提取出所述脑部医疗影像集中所述脑组织区域为浅灰色、所述脑脊液及所述脑室为黑色、所述外侧颅骨为白色的图片集。
  6. 如权利要求4所述的脑出血点智能检测方法，其中，所述原始脑部出血点检测模型是基于卷积神经网络模型，所述模型参数包括对比度、亮度、滚动相邻的脑部图像切片和出血点区域的置信度。
  7. 如权利要求6所述的脑出血点智能检测方法，其中，所述卷积神经网络模型包括卷积操作和池化操作，所述卷积操作和池化操作包括：
    预先构建卷积模板并确定卷积步长；
    根据所述卷积步长，将卷积模板与所述训练集进行计算得到卷积操作后的卷积矩阵集，完成所述卷积操作；
    选择所述卷积矩阵集中矩阵的最大值或平均值替代所述卷积矩阵集，完成所述池化操作。
  8. 一种脑出血点智能检测装置，其中，所述装置包括：
    数据接收及处理模块：用于获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
    图像调整模块：用于基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
    出血点检测训练模型模块：用于将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
    出血点结果输出模块：接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
  9. 一种电子设备，其中，所述电子设备包括存储器和处理器，所述存储器中存储有可在所述处理器上运行的脑出血点智能检测程序，其中，所述脑出血点智能检测程序被所述处理器执行时实现如下步骤：
    获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
    基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集；
    将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
    接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
  10. 如权利要求9所述的电子设备，其中，所述格式转换的计算方法为：
    Figure PCTCN2020131904-appb-100002
    其中，v为所述脑部医疗影像集的图像存储值，G为计算机操作平台上的显示值，gm为计算机显示器显示的最大位图文件为255，w为所述脑部医疗影像集的窗宽，c为所述脑部医疗影像集的窗位。
  11. 如权利要求9所述的电子设备，其中，所述参数调整算法包括向量生成函数、图像提取函数、图像显示函数和灰度拉伸函数；
    所述基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集，包括：
    基于所述向量生成函数，将所述原始可读脑部图像集的像素数据存储到无符号整型数组，并将所述无符号整型数组转变为向量集；
    根据所述图像提取函数从所述向量集中提取脑出血图像数据，并基于所述图像显示函数进行图像显示；
    根据所述图像显示结果，调用所述灰度拉伸函数进行参数调节得到调节后的可读脑部图像集。
  12. 如权利要求9所述的电子设备，其中，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，包括：
    将所述训练集进行脑部出血点标记，得到真实出血点标记坐标集；
    初始化所述原始脑部出血点检测模型的模型参数；
    将所述训练集输入至所述原始脑部出血点检测模型中，所述原始脑部出血点检测模型根据所述训练集不断调整所述模型参数并输出预测出血点标记坐标集；
    判断所述预测出血点标记坐标集与所述真实出血点标记坐标集的误差是否小于预设误差，若大于所述预设误差，所述原始脑部出血点检测模型继续调整所述模型参数，若小于所述预设误差，完成所述训练得到所述脑部出血点检测模型。
  13. 如权利要求9至12中任意一项所述的电子设备，其中，所述提纯操作包括：
    自动化标定所述脑部医疗影像集中脑组织区域、脑脊液、脑室及外侧颅骨；
    提取出所述脑部医疗影像集中所述脑组织区域为浅灰色、所述脑脊液及所述脑室为黑色、所述外侧颅骨为白色的图片集。
  14. 如权利要求12所述的电子设备，其中，所述原始脑部出血点检测模型是基于卷积神经网络模型，所述模型参数包括对比度、亮度、滚动相邻的脑部图像切片和出血点区域的置信度。
  15. 一种计算机可读存储介质，其中，所述计算机可读存储介质上存储有脑出血点智能检测程序，所述脑出血点智能检测程序可被一个或者多个处理器执行，以实现如下步骤：
    获取脑部医疗影像集，并基于脑出血的性质对所述脑部医疗影像集进行提纯操作和格式转换，生成原始可读脑部图像集；
    基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部 图像集；
    将所述可读脑部图像集分为训练集和测试集，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，将所述测试集输入至所述脑部出血点检测模型得到测试值，若所述测试值小于预设阈值，所述脑部出血点检测模型接收所述训练集继续训练，若所述测试值大于预设阈值，所述脑部出血点检测模型完成所述训练；
    接收用户输入的脑部图片，将所述脑部图片输入至所述脑部出血点检测模型得到脑部出血点坐标，基于预先构建的出血点操作平台和所述脑部出血点坐标进行编辑整理输出脑部出血点。
  16. 如权利要求15所述的计算机可读存储介质，其中，所述格式转换的计算方法为：
    Figure PCTCN2020131904-appb-100003
    其中，v为所述脑部医疗影像集的图像存储值，G为计算机操作平台上的显示值，gm为计算机显示器显示的最大位图文件为255，w为所述脑部医疗影像集的窗宽，c为所述脑部医疗影像集的窗位。
  17. 如权利要求15所述的计算机可读存储介质，其中，所述参数调整算法包括向量生成函数、图像提取函数、图像显示函数和灰度拉伸函数；
    所述基于预先构建的参数调整算法调整所述原始可读脑部图像集得到可读脑部图像集，包括：
    基于所述向量生成函数，将所述原始可读脑部图像集的像素数据存储到无符号整型数组，并将所述无符号整型数组转变为向量集；
    根据所述图像提取函数从所述向量集中提取脑出血图像数据，并基于所述图像显示函数进行图像显示；
    根据所述图像显示结果，调用所述灰度拉伸函数进行参数调节得到调节后的可读脑部图像集。
  18. 如权利要求15所述的计算机可读存储介质，其中，将所述训练集输入至预先构建的原始脑部出血点检测模型中训练得到脑部出血点检测模型，包括：
    将所述训练集进行脑部出血点标记，得到真实出血点标记坐标集；
    初始化所述原始脑部出血点检测模型的模型参数；
    将所述训练集输入至所述原始脑部出血点检测模型中，所述原始脑部出血点检测模型根据所述训练集不断调整所述模型参数并输出预测出血点标记坐标集；
    判断所述预测出血点标记坐标集与所述真实出血点标记坐标集的误差是否小于预设误差，若大于所述预设误差，所述原始脑部出血点检测模型继续调整所述模型参数，若小于所述预设误差，完成所述训练得到所述脑部出血点检测模型。
  19. 如权利要求15至18中任意一项所述的计算机可读存储介质，其中，所述提纯操作包括：
    自动化标定所述脑部医疗影像集中脑组织区域、脑脊液、脑室及外侧颅骨；
    提取出所述脑部医疗影像集中所述脑组织区域为浅灰色、所述脑脊液及所述脑室为黑色、所述外侧颅骨为白色的图片集。
  20. 如权利要求18所述的计算机可读存储介质，其中，所述原始脑部出血点检测模型是基于卷积神经网络模型，所述模型参数包括对比度、亮度、滚动相邻的脑部图像切片和出血点区域的置信度。
PCT/CN2020/131904 2019-12-09 2020-11-26 脑出血点智能检测方法、装置、电子设备及存储介质 WO2021115130A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911254426.9 2019-12-09
CN201911254426.9A CN111179222B (zh) 2019-12-09 2019-12-09 脑出血点智能检测方法、装置及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2021115130A1 true WO2021115130A1 (zh) 2021-06-17

Family

ID=70646253

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/131904 WO2021115130A1 (zh) 2019-12-09 2020-11-26 脑出血点智能检测方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN111179222B (zh)
WO (1) WO2021115130A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116630247A (zh) * 2023-05-06 2023-08-22 河北省儿童医院(河北省第五人民医院、河北省儿科研究所) 脑血流图像处理方法及装置、脑血流监测系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111179222B (zh) * 2019-12-09 2023-08-22 平安科技（深圳）有限公司 脑出血点智能检测方法、装置及计算机可读存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160063726A1 (en) * 2014-08-28 2016-03-03 Koninklijke Philips N.V. Model-based segmentation of an anatomical structure
CN109190690A (zh) * 2018-08-17 2019-01-11 东北大学 基于机器学习的swi图像的脑微出血点检测识别方法
CN110503630A (zh) * 2019-07-19 2019-11-26 江苏师范大学 一种基于三维深度学习模型的脑出血分类、定位与预测方法
CN111179222A (zh) * 2019-12-09 2020-05-19 平安科技（深圳）有限公司 脑出血点智能检测方法、装置及计算机可读存储介质

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106621040B (zh) * 2016-12-21 2023-07-04 北京品驰医疗设备有限公司 一种脑深部电刺激出血检测系统
CN108898583B (zh) * 2018-06-12 2019-08-13 东软医疗系统股份有限公司 微出血点的检测方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160063726A1 (en) * 2014-08-28 2016-03-03 Koninklijke Philips N.V. Model-based segmentation of an anatomical structure
CN109190690A (zh) * 2018-08-17 2019-01-11 东北大学 基于机器学习的swi图像的脑微出血点检测识别方法
CN110503630A (zh) * 2019-07-19 2019-11-26 江苏师范大学 一种基于三维深度学习模型的脑出血分类、定位与预测方法
CN111179222A (zh) * 2019-12-09 2020-05-19 平安科技（深圳）有限公司 脑出血点智能检测方法、装置及计算机可读存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116630247A (zh) * 2023-05-06 2023-08-22 河北省儿童医院(河北省第五人民医院、河北省儿科研究所) 脑血流图像处理方法及装置、脑血流监测系统
CN116630247B (zh) * 2023-05-06 2023-10-20 河北省儿童医院(河北省第五人民医院、河北省儿科研究所) 脑血流图像处理方法及装置、脑血流监测系统

Also Published As

Publication number Publication date
CN111179222B (zh) 2023-08-22
CN111179222A (zh) 2020-05-19

Similar Documents

Publication Publication Date Title
JP7058373B2 (ja) 医療画像に対する病変の検出及び位置決め方法、装置、デバイス、及び記憶媒体
WO2021189912A1 (zh) 图像中目标物的检测方法、装置、电子设备及存储介质
CN110428475B (zh) 一种医学图像的分类方法、模型训练方法和服务器
US10789504B2 (en) Method and device for extracting information in histogram
WO2019109526A1 (zh) 人脸图像的年龄识别方法、装置及存储介质
WO2019218451A1 (zh) 一种医学报告的生成方法及设备
WO2022247005A1 (zh) 图像中目标物识别方法、装置、电子设备及存储介质
WO2021012494A1 (zh) 基于深度学习的人脸识别方法、装置及计算机可读存储介质
WO2020103417A1 (zh) 一种bmi评测方法、装置及计算机可读存储介质
WO2020253508A1 (zh) 异常细胞检测方法、装置及计算机可读存储介质
WO2018057714A1 (en) Systems, methods and media for automatically generating a bone age assessment from a radiograph
CN108154509B (zh) 癌症识别方法、装置及存储介质
CN109003269B (zh) 一种能提高医生效率的医学影像病灶的标注提取方法
CN111986785B (zh) 医学影像标注方法和装置、设备及存储介质
WO2021115130A1 (zh) 脑出血点智能检测方法、装置、电子设备及存储介质
US20200019823A1 (en) Medical image analysis method applying machine learning and system thereof
WO2019223147A1 (zh) 肝脏癌变定位方法、装置及存储介质
WO2021151338A1 (zh) 医学影像图片分析方法、装置、电子设备及可读存储介质
WO2021147219A1 (zh) 基于图像的文本识别方法、装置、电子设备及存储介质
JP2023539483A (ja) 医用画像処理方法、装置、機器、記憶媒体及びコンピュータプログラム
WO2021189856A1 (zh) 证件校验方法、装置、电子设备及介质
WO2021184576A1 (zh) 医学图像的生成方法、装置、电子设备及介质
WO2021190664A1 (zh) 基于关键点定位的多人脸检测方法、系统及存储介质
US20200342276A1 (en) Normalization method for machine-learning and apparatus thereof
WO2020248848A1 (zh) 智能化异常细胞判断方法、装置及计算机可读存储介质

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: 20898279

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: 20898279

Country of ref document: EP

Kind code of ref document: A1