CN109259784A - AI prediction technique, device, equipment and the storage medium of cerebral infarction - Google Patents
AI prediction technique, device, equipment and the storage medium of cerebral infarction Download PDFInfo
- Publication number
- CN109259784A CN109259784A CN201810983952.8A CN201810983952A CN109259784A CN 109259784 A CN109259784 A CN 109259784A CN 201810983952 A CN201810983952 A CN 201810983952A CN 109259784 A CN109259784 A CN 109259784A
- Authority
- CN
- China
- Prior art keywords
- cerebral infarction
- image
- training
- convolutional neural
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 206010008118 cerebral infarction Diseases 0.000 title claims abstract description 126
- 208000026106 cerebrovascular disease Diseases 0.000 title claims abstract description 126
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000003860 storage Methods 0.000 title claims description 30
- 238000012549 training Methods 0.000 claims abstract description 116
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 93
- 238000012360 testing method Methods 0.000 claims abstract description 23
- 238000004590 computer program Methods 0.000 claims description 24
- 238000000605 extraction Methods 0.000 claims description 10
- 230000008439 repair process Effects 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 5
- 230000001537 neural effect Effects 0.000 claims description 4
- 230000008901 benefit Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 210000004218 nerve net Anatomy 0.000 claims 1
- 230000006870 function Effects 0.000 description 21
- 238000010586 diagram Methods 0.000 description 13
- 238000012545 processing Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 11
- 230000004913 activation Effects 0.000 description 8
- 210000002569 neuron Anatomy 0.000 description 8
- 238000013507 mapping Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 210000004556 brain Anatomy 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 5
- 230000005291 magnetic effect Effects 0.000 description 5
- 230000011218 segmentation Effects 0.000 description 5
- 206010061216 Infarction Diseases 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000007574 infarction Effects 0.000 description 4
- 230000003902 lesion Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 239000008280 blood Substances 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 230000002490 cerebral effect Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000002583 angiography Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000000302 ischemic effect Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 206010008088 Cerebral artery embolism Diseases 0.000 description 1
- 206010008092 Cerebral artery thrombosis Diseases 0.000 description 1
- 206010008132 Cerebral thrombosis Diseases 0.000 description 1
- 206010008190 Cerebrovascular accident Diseases 0.000 description 1
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 description 1
- 201000001429 Intracranial Thrombosis Diseases 0.000 description 1
- 208000032382 Ischaemic stroke Diseases 0.000 description 1
- 208000004552 Lacunar Stroke Diseases 0.000 description 1
- 206010051078 Lacunar infarction Diseases 0.000 description 1
- 238000005481 NMR spectroscopy Methods 0.000 description 1
- 206010028851 Necrosis Diseases 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 208000031481 Pathologic Constriction Diseases 0.000 description 1
- 208000006011 Stroke Diseases 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000036770 blood supply Effects 0.000 description 1
- 210000005013 brain tissue Anatomy 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005345 coagulation Methods 0.000 description 1
- 230000015271 coagulation Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000008103 glucose Substances 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 201000010849 intracranial embolism Diseases 0.000 description 1
- 230000003907 kidney function Effects 0.000 description 1
- 150000002632 lipids Chemical class 0.000 description 1
- 230000003908 liver function Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000017074 necrotic cell death Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000036262 stenosis Effects 0.000 description 1
- 208000037804 stenosis Diseases 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
- 210000002700 urine Anatomy 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
- A61B6/03—Computed tomography [CT]
- A61B6/032—Transmission computed tomography [CT]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/501—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of the head, e.g. neuroimaging or craniography
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5205—Devices using data or image processing specially adapted for radiation diagnosis involving processing of raw data to produce diagnostic data
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Surgery (AREA)
- High Energy & Nuclear Physics (AREA)
- Physics & Mathematics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Optics & Photonics (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Neurology (AREA)
- Neurosurgery (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Pulmonology (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of AI prediction techniques of cerebral infarction, first acquisition training sample；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；Then the convolutional neural networks based on coding plus decoding architecture that training is completed are obtained；Wherein, the convolutional neural networks based on coding plus decoding architecture that the training is completed are by carrying out convolutional neural networks training to the training sample and initial convolutional neural networks, prediction result is assessed using dice coefficient when training, backpropagation is carried out to the initial convolutional network according to the assessment result；CT image to be predicted is pre-processed, test sample is obtained；The convolutional neural networks based on coding plus decoding architecture completed according to the training carry out cerebral infarction prediction to the CT image to be predicted, obtain prediction result.By the image recognition of the convolutional neural networks based on coding plus decoding architecture, the region for whether having cerebral infarction and cerebral infarction can be predicted.
Description
Technical field
The present invention relates to technical field of information processing more particularly to a kind of AI prediction technique of cerebral infarction, device, equipment and
Storage medium.
Background technique
Cerebral infarction is once called as cerebral infarction, also known as cerebral arterial thrombosis (cerebral ischemic stroke), refers to because of brain
Portion's blood supply obstacle, the ischemic necrosis or softening of limitation brain tissue caused by ischemic, anoxic.The clinic of cerebral infarction is often
See that type has cerebral thrombosis, lacunar infarction and cerebral embolism etc., cerebral infarction accounts for the 80% of whole cerebral apoplexies, and cerebral infarction is not only given
Human health and life cause great threat, and bring greatly pain and heavy burden to patient, family and society.
Currently, the recognition methods to cerebral infarction, typically goes to hospital to carry out CT examination, MRI inspection, routine inspection or spy
It is different to check；When carrying out CT examination, CT shows that infarct is low-density, can specify position, the form and dimension of lesion, larger
Infarct the ventricles of the brain can be made to be pressurized, deformation and centerline construction displacement, but in cerebral infarction onset 4~6 hours, only some cases can
See the slightly lower density stove of obscure boundary, and most case could show the more clear low-density stove in boundary after 24 hours, and
Infarct less than 5mm.MRI inspection is exactly nuclear magnetic resonance image check.Routine inspection is exactly blood, urine, stool routine examination and liver function
Energy, renal function, coagulation function, blood glucose, blood lipid, electrocardiogram etc. are used as routine inspection, special examined such as transcranial Doppler ultrasound
(TCD), arteria carotis Color B-Type Ultrasonic, magnetic resonance, angiography (MRA), Digital Subtraction aortocranial angiography (DSA), arteria carotis are made
Shadow, can define that whether there is or not Intra-extracranial artery stenosis or occlusions.
In implementing the embodiments of the present invention, discovery there is not one kind that can carry out cerebral infarction AI prediction to inventor in the prior art
Technology, because cerebral infarction morbidity is more anxious, more absence of prodrome, focal neurosigns peak at several minutes to a few hours, and
And multilist shows complete stroke, such as without a kind of technology that can be predicted with cerebral infarction, then can not provide in time relief to patient.
Summary of the invention
In view of this, the embodiment of the present invention provides AI prediction technique, device, equipment and the storage medium of a kind of cerebral infarction,
Whether by the image recognition technology of convolutional neural networks, can predict has cerebral infarction, so that patient is succoured in time.
In a first aspect, including the following steps: the embodiment of the invention provides a kind of AI prediction technique of cerebral infarction
Obtain training sample；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；
Obtain the convolutional neural networks based on coding plus decoding architecture that training is completed；Wherein, the base that the training is completed
In the convolutional neural networks of coding plus decoding architecture by carrying out convolution mind to the training sample and initial convolutional neural networks
Through network training, prediction result is assessed using dice coefficient when training, according to the assessment result to the initial volume
Product network carries out backpropagation.
CT image to be predicted is pre-processed, test sample is obtained；
The convolutional neural networks based on coding plus decoding architecture completed according to the training, scheme the CT to be predicted
As carrying out cerebral infarction prediction, prediction result is obtained.
In the first possible implementation of first aspect, described obtain trains that completes to add decoding architecture based on coding
Convolutional neural networks；Wherein, the convolutional neural networks based on coding plus decoding architecture that the training is completed pass through to described
Training sample and initial convolutional neural networks carry out convolutional neural networks training, when training using dice coefficient to prediction result into
Row assessment carries out backpropagation to the initial convolutional network according to the assessment result, comprising:
Obtain initial convolutional neural networks；
The original CT image is encoded based on the initial convolutional neural networks, extracted by convolution operation described in
The corresponding characteristic pattern of each layer of original CT image；
The characteristic pattern is decoded based on the initial convolutional neural networks, the characteristic pattern after handling the extraction,
And cerebral infarction focal zone is predicted；
Prediction result assessment, using prediction result set with the cross entropy of the corresponding cerebral infarction tag set as prediction
Penalty values, and solved by dice coefficient, to realize backpropagation.
With reference to first aspect and the first possible mode of first aspect, second in first aspect can the side of being able to achieve
It is described that the original CT image is encoded based on the initial convolutional neural networks in formula, institute is extracted by convolution operation
Stating the corresponding characteristic pattern of each layer of original CT image includes:
Based on Mobilenet framework and porous spatial pyramid pond module, it is right to extract each layer of the original CT image
The characteristic pattern answered.
The possible implementation of with reference to first aspect the first, in the third possible implementation of first aspect, institute
It states and the characteristic pattern is decoded based on the initial convolutional neural networks, the characteristic pattern after handling the extraction, and to brain
It predicts in infarction lesion area
When handling the characteristic pattern after the extraction, the preceding N layers of initial corresponding characteristic pattern obtained by coding are added,
To repair the loss of cut zone boundary information；Wherein, N is greater than 1.
It is described CT image to be predicted to be pre-processed in the 4th kind of possible implementation of first aspect, it obtains
Include: to test sample
The pixel value of the CT image to be predicted is normalized into preset section, obtains test sample.
In the 5th kind of possible implementation of first aspect, the acquisition training sample；Wherein, the training sample packet
It includes original CT image and corresponding cerebral infarction label includes:
Obtain the original CT image collected；
Obtain the label that cerebral infarction judgement is carried out to each original CT image.
In the 6th kind of possible implementation of first aspect, the acquisition doctor carries out each original CT image
Cerebral infarction judgement label include:
To each original CT image, when judging the original CT image to there is cerebral infarction, acquisition has cerebral infarction mark
Label and corresponding cerebral infarction region；
To each original CT image, when judging the original CT image there is no cerebral infarction, obtain without cerebral infarction
Label.
Second aspect, the embodiment of the invention also provides a kind of AI prediction meanss of cerebral infarction, comprising:
Training sample obtains module, for obtaining training sample；Wherein, the training sample includes original CT image and right
The cerebral infarction label answered；
Network obtains module, for obtaining the convolutional neural networks based on coding plus decoding architecture of training completion；Wherein,
The convolutional neural networks based on coding plus decoding architecture that the training is completed pass through to the training sample and initial convolution mind
Convolutional neural networks training is carried out through network, prediction result is assessed using dice coefficient when training, according to the assessment
As a result backpropagation is carried out to the initial convolutional network.
Test sample obtains module and obtains test sample for pre-processing to CT image to be predicted；
Prediction module, the convolutional neural networks based on coding plus decoding architecture for being completed according to the training, to institute
It states CT image to be predicted and carries out cerebral infarction prediction, obtain prediction result.
The third aspect, the embodiment of the invention also provides a kind of pre- measurement equipments of the AI of cerebral infarction, which is characterized in that including place
It manages device, memory and storage in the memory and is configured as the computer program executed by the processor, it is described
Processor realizes the AI prediction technique of cerebral infarction as described above when executing the computer program.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, which is characterized in that the meter
Calculation machine readable storage medium storing program for executing includes the computer program of storage, wherein controls the calculating in computer program operation
Equipment where machine readable storage medium storing program for executing executes the AI prediction technique of cerebral infarction as described above.
Implement the embodiment of the present invention to have the following beneficial effects:
Training sample is obtained first；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；So
The convolutional neural networks based on coding plus decoding architecture that training is completed are obtained afterwards；Wherein, it is described training complete based on coding
Add the convolutional neural networks of decoding architecture by carrying out convolutional neural networks to the training sample and initial convolutional neural networks
Training is assessed prediction result using dice coefficient when training, according to the assessment result to the initial convolutional network
Carry out backpropagation.In the training process, using the original CT image and corresponding cerebral infarction label to initial convolutional Neural
Network training, and default result and corresponding cerebral infarction label are assessed by dice coefficient, backpropagation is carried out to reach
To optimal convolutional neural networks parameter；CT image to be predicted is pre-processed, test sample is obtained；According to the training
The convolutional neural networks based on coding plus decoding architecture completed carry out cerebral infarction prediction to the CT image to be predicted, obtain
To prediction result.By it is described based on coding plus decoding architecture convolutional neural networks image recognition, can predict whether
There is the region of cerebral infarction and cerebral infarction.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, of the invention other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is the schematic diagram of the pre- measurement equipment of AI for the cerebral infarction that the embodiment of the present invention one provides；
Fig. 2 is a kind of flow diagram of the AI prediction technique of cerebral infarction provided by Embodiment 2 of the present invention；
Fig. 3 is the schematic diagram of the AI prediction of cerebral infarction provided by Embodiment 2 of the present invention；
Fig. 4 is the structural schematic diagram of the AI prediction meanss for the cerebral infarction that third embodiment of the invention provides.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Those skilled in the art will appreciate that the present invention can be implemented as equipment, method or computer program product.
Therefore, the present disclosure may be embodied in the following forms, it may be assumed that can be complete hardware, be also possible to complete software (including
Firmware, resident software, microcode etc.), it can also be the form that hardware and software combines, referred to generally herein as " circuit ", " mould
Block " or " system ".In addition, in some embodiments, the present invention is also implemented as in one or more computer-readable mediums
In computer program product form, include computer-readable program code in the computer-readable medium.
It can be using any combination of one or more computer-readable media.Computer-readable medium can be calculating
Machine readable signal medium or computer readable storage medium.Computer readable storage medium for example can be --- but it is unlimited
In system, device or the device of --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or any above combination.It calculates
The more specific example (non exhaustive list) of machine readable storage medium storing program for executing includes: electrical connection with one or more conducting wires, just
Taking formula computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable type may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In this document, computer readable storage medium can be it is any include or storage journey
The tangible medium of sequence, the program can be commanded execution system, device or device use or in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including --- but
It is not limited to --- electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be
Any computer-readable medium other than computer readable storage medium, which can send, propagate or
Transmission is for by the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In --- wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on a remote computer or server completely on the remote computer on the user computer.?
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as mentioned using Internet service
It is connected for quotient by internet).
Below with reference to the method for the embodiment of the present invention, the flow chart of equipment (system) and computer program product and/or
The block diagram description present invention.It should be appreciated that each box in each box and flowchart and or block diagram of flowchart and or block diagram
Combination, can be realized by computer program instructions.These computer program instructions can be supplied to general purpose computer, dedicated
The processor of computer or other programmable data processing units, to produce a kind of virtual machine, these computer programs refer to
It enables and being executed by computer or other programmable data processing units, produced in the box in implementation flow chart and/or block diagram
The device of defined function/operation.
These computer program instructions can also be stored in can make computer or other programmable data processing units
In computer-readable medium operate in a specific manner, in this way, the instruction of storage in computer-readable medium just produces one
Function/operation command device specified in a box including in implementation flow chart and/or block diagram
(instructionmeans) manufacture (manufacture).
Computer program instructions can also be loaded into computer, other programmable data processing units or other equipment
On, so that series of operation steps are executed in computer, other programmable data processing units or other equipment, in terms of generating
The process that calculation machine is realized, so that the instruction executed on a computer or other programmable device is capable of providing implementation flow chart
And/or function/operation process specified in the box in block diagram.
Embodiment one
Referring to Figure 1, Fig. 1 is the schematic diagram of the pre- measurement equipment of AI for the cerebral infarction that the embodiment of the present invention one provides, for holding
The AI prediction technique of row cerebral infarction provided in an embodiment of the present invention, as shown in Figure 1, the pre- measurement equipment of the AI of the cerebral infarction include: to
A few processor 11, such as CPU, at least one network interface 14 or other users interface 13, memory 15, at least one
Communication bus 12, communication bus 12 is for realizing the connection communication between these components.Wherein, user interface 13 optionally can be with
Including USB interface and other standards interface, wireline interface.Network interface 14 optionally may include Wi-Fi interface and its
His wireless interface.Memory 15 may include high speed RAM memory, it is also possible to further include non-labile memory (non-
Volatilememory), a for example, at least magnetic disk storage.Memory 15 optionally may include at least one and be located remotely from
The storage device of aforementioned processor 11.
In some embodiments, memory 15 stores following element, executable modules or data structures, or
Their subset or their superset:
Operating system 151 includes various system programs, for realizing various basic businesses and hardware based of processing
Business；
Program 152.
Specifically, processor 11 executes brain described in above-described embodiment for calling the program 152 stored in memory 15
The AI prediction technique of infraction.
Alleged processor can be central processing unit (Central Processing Unit, CPU), can also be it
His general processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng the processor is the control centre of the AI prediction technique of the cerebral infarction, utilizes the entire institute of various interfaces and connection
State the various pieces of the AI prediction technique of cerebral infarction.
The memory can be used for storing the computer program and/or module, and the processor is by operation or executes
Computer program in the memory and/or module are stored, and calls the data being stored in memory, realizes cerebral infarction
The various functions of the electronic device of the AI prediction of plug.The memory can mainly include storing program area and storage data area,
In, storing program area can application program needed for storage program area, at least one function (such as sound-playing function, text
Conversion function etc.) etc.；Storage data area can store that (for example audio data, text disappear according to the created data that use of mobile phone
Cease data etc.) etc..In addition, memory may include high-speed random access memory, it can also include nonvolatile memory, example
Such as hard disk, memory, plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card), at least one disk memory, flush memory device or other volatibility are solid
State memory device.
Wherein, if the module that the AI prediction of the cerebral infarction integrates is realized in the form of SFU software functional unit and as only
Vertical product when selling or using, can store in a computer readable storage medium.Based on this understanding, this hair
All or part of the process in bright realization above-described embodiment method, can also be instructed by computer program relevant hardware come
It completes, the computer program can be stored in a computer readable storage medium, and the computer program is by processor
When execution, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program code,
The computer program code can be source code form, object identification code form, executable file or certain intermediate forms etc..Institute
State computer-readable medium may include: can carry the computer program code any entity or device, recording medium,
USB flash disk, mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), arbitrary access
Memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..It needs
It is bright, the content that the computer-readable medium includes can according in jurisdiction make laws and patent practice requirement into
Row increase and decrease appropriate, such as do not include electric load according to legislation and patent practice, computer-readable medium in certain jurisdictions
Wave signal and telecommunication signal.
Below with reference to accompanying drawings come describe the embodiment of the present invention cerebral infarction AI prediction method.
Embodiment two
Fig. 2 is a kind of flow diagram of the AI prediction technique of cerebral infarction provided by Embodiment 2 of the present invention.
A kind of AI prediction technique of cerebral infarction, comprising the following steps:
S11, training sample is obtained；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；
S12, the convolutional neural networks based on coding plus decoding architecture that training is completed are obtained；Wherein, the training is completed
Based on coding plus decoding architecture convolutional neural networks by being rolled up to the training sample and initial convolutional neural networks
Product neural metwork training, assesses prediction result using dice coefficient when training, according to the assessment result to described first
Beginning convolutional network carries out backpropagation.
S13, CT image to be predicted is pre-processed, obtains test sample；
S14, the convolutional neural networks based on coding plus decoding architecture completed according to the training, to described to be predicted
CT image carries out cerebral infarction prediction, obtains prediction result.
It should be noted that having many linear methods, such as linear regression in numerous image recognition technologys, logarithm is several
Rate returns.But under nonlinear prediction scene, linear regression is not available then, it is necessary to be taken nonlinear regression, such as be supported
Vector machine, there are also famous neural networks.The cardinal principle of neural network is to imitate the working principle of cerebral neuron, by one
A neuron connects networking, and each neuron performs some processing the signal of input, then will treated signal spreads out of work
For the input of target nerve member.Processing may be cumulative, it may be possible to filtering or any mode.It is simply advised by single neuron
Extensive repetition then forms a complication system, is finally reached the purpose of nonlinear prediction.Network treats recognition mode in advance
Great amount of samples carry out learning training, and by the interconnection weight between neuron come storage result, network training is finished, mark
It there is the ability correctly classified to certain a kind of or several quasi-mode, in this way, once there is new mode sample to be identified to occur,
Network just at once can make " reflection " it according to memory, and identification.The basic structure of convolutional neural networks includes two layers, one
For convolution operation, the input of each neuron is connected with the local acceptance region of preceding layer, extracts the part by convolution operation
Feature.After the local feature is extracted, its positional relationship between other feature is also decided therewith；The second is activation
Each computation layer of function, network is made of multiple Feature Mappings, and each Feature Mapping is a plane.Feature Mapping structure is adopted
Use relu function as the activation primitive of convolutional network.Further, since the neuron on a mapping face shares weight, thus subtract
The number of network freedom parameter is lacked.In each of general convolutional neural networks convolutional layer all followed by pondization behaviour
Make, compressed characteristic pattern is obtained by pond and reduces feature resolution.
Preferably, the convolutional neural networks based on coding plus decoding architecture for obtaining training and completing；Wherein, the instruction
Practice the convolutional neural networks based on coding plus decoding architecture completed to pass through to the training sample and initial convolutional neural networks
Convolutional neural networks training is carried out, prediction result is assessed using dice coefficient when training, according to the assessment result pair
The initial convolutional network carries out backpropagation, comprising:
Obtain initial convolutional neural networks；
The original CT image is encoded based on the initial convolutional neural networks, extracted by convolution operation described in
The corresponding characteristic pattern of each layer of original CT image；
The characteristic pattern is decoded based on the initial convolutional neural networks, the characteristic pattern after handling the extraction,
And cerebral infarction focal zone is predicted；
Prediction result assessment, using prediction result set with the cross entropy of the corresponding cerebral infarction tag set as prediction
Penalty values, and solved by dice coefficient, to realize backpropagation.
In the present embodiment, the convolutional neural networks (CNN) are by input layer, convolutional layer, activation primitive, pond layer, complete
Articulamentum composition.Convolutional layer is used to carry out feature extraction, i.e. convolution operation, and input picture is 32*32*3,3 be its depth (i.e.
R, G, B), wherein rgb color mode is a kind of color standard of industry, is by red (R), green (G), blue (B) three face
To obtain miscellaneous color, RGB is to represent red, green, blue for the variation of chrominance channel and their mutual superpositions
The color in three channels.Convolutional layer is the filter (receptive field) of a 5*5*3, it should be noted that: the depth of receptive field must be with
The depth of input picture is identical.Pass through the characteristic pattern of the available 28*28*1 of the convolution of a filter and input picture.
Deeper characteristic pattern would generally be obtained using multilayer convolutional layer.The process of convolution is as follows: input picture and filter's
Corresponding position element multiplication is summed again, finally adds B, obtains characteristic pattern, and convolutional layer is exactly that " weight is total there are one characteristic
Enjoy " principle, so-called weight it is shared that is, give an input picture, gone to sweep this figure, the inside filter with a filter
Number be just weight, this schemes each position and is swept by same filter, so weight is the same, that is, total
It enjoys.Activation primitive: if input vary less, cause export structure occur it is completely different as a result, such case be we not
It wishes to, in order to simulate subtleer variation, outputs and inputs numerical value and be more than 0 to 1, can be any between 0 and 1
Number, activation primitive is for non-linear factor is added, because the expressiveness of linear model is inadequate.For image, we are main
It is handled by the way of convolution, that is, assigns a weight to each pixel, this operation is obviously exactly linear.
But for our samples, it is not necessarily linear separability, in order to solve this problem, we can linearly be become
Change or we introduce non-linear factor, solves the problems, such as that linear model cann't be solved.Because of the Fundamentals of Mathematics of neural network
Being everywhere can be micro-, so the activation primitive chosen will can guarantee data input and output and can be micro-, operating characteristic is not
It is disconnected to carry out cycle calculations, so the value of each neuron is also continually changing in per generation cyclic process.Which results in
Effect when tanh feature differs obvious can constantly can expand characteristic effect in cyclic process and show very well, but, in spy
When sign difference is more complicated or difference is not especially big, when needing subtleer classification to judge, sigmoid effect is all right
?.Sigmoid and tanh are being used as activation primitive, centainly it is noted that having to carry out normalizing words to input, otherwise
Value after activation can all enter flat region, keep the output of hidden layer all convergent, but Relu does not need input and is normalized to prevent
Only they reach saturation.Pond layer: compressing the characteristic pattern of input, on the one hand characteristic pattern made to become smaller, and simplifies network query function
Complexity；On the one hand Feature Compression is carried out, main feature is extracted.Pond layer often behind convolutional layer, is reduced by pond
The feature vector of convolutional layer output, while improving result (being less prone to over-fitting).Because image has a kind of " nature static "
Attribute, this also means that very likely equally applicable in another region in an image-region useful feature.Therefore, it is
Description big image, one naturally idea be exactly that aggregate statistics are carried out to the feature of different location, for example, people can be with
The average value (or maximum value) of some special characteristic on one region of image is calculated to represent the feature in this region.Porous air
Between pyramid pondization the convolution feature of the image of any scale can be converted to identical dimensional, this can not only allow CNN to handle
The image of any scale, moreover it is possible to avoid cropping and warping from operating, lead to the loss of some information, have extremely important
Meaning.The size that general CNN requires input picture is fixed, this is because the input of full articulamentum need to fix it is defeated
Enter dimension, but in convolution operation be it is restricted not to graphical rule, in the present embodiment, propose porous spatial pyramid pond
Change, first allows image to carry out convolution operation, be then converted into the identical feature of dimension and be input to full articulamentum, this can be CNN
Expand to the image of arbitrary size.Full articulamentum connects all features, and giving output valve to classifier, (such as softmax classifies
Device).
Further, described that the original CT image is encoded based on the initial convolutional neural networks, pass through volume
The corresponding characteristic pattern of each layer of original CT image is extracted in product operation
Based on Mobilenet framework and porous spatial pyramid pond module, it is right to extract each layer of the original CT image
The characteristic pattern answered.
Further, described that the characteristic pattern is decoded based on the initial convolutional neural networks, it is mentioned described in processing
Characteristic pattern after taking, and include: to the prediction of cerebral infarction focal zone
When handling the characteristic pattern after the extraction, the preceding N layers of initial corresponding characteristic pattern obtained by coding are added,
To repair the loss of cut zone boundary information；Wherein, N is greater than 1.
Specifically, in the present embodiment, it is described original using the convolutional neural networks based on coding plus decoding architecture
CT image extracts feature by coded portion, and then used decoded portion is to handle the feature after extracting and to cerebral infarction lesion
It is predicted in area.Mobilenet framework and porous spatial pyramid pond module are used in coded portion network, it is described
Mobilenet construction characteristic is to replace general convolution operation to reduce network parameter, the porous air using the revoluble product of depth
Between pyramid pond module the advantages of be that can repair the loss of cut zone boundary information with reference to the feature of different scale.?
In decoded portion by be added in coded portion it is more original before several layers of Feature Mappings repair cut zone boundary information
It loses.The cross entropy of one prediction result of generation after prediction, prediction result collection and the cerebral infarction label is as prediction
Penalty values, optimal result cross entropy are 0.Regard each layer network as a function about input CT images data, then
Cross entropy equally can be regarded as a function of network parameter, and the accuracy of prediction can be made by the minimum value found a function
Reach highest.Function is minimized corresponding parameter then and is the parameter of target network.What solution procedure can be declined by gradient
Mode carries out.This process is referred to as backpropagation, by backpropagation, so that penalty values reach minimum to obtain optimal net
Network parameter.
In the present embodiment, the Dice coefficient is a kind of set measuring similarity function, and value range is (0~1), is led to
It is usually used in calculating the similarity of two samples:
It uses dice coefficient as the index of AI lesion segmentation result in the present system, segmentation is indicated when dice coefficient is 0
As a result segmentation result and actual focal region perfect registration are indicated when dice coefficient is 1 without being overlapped with actual focal region.
In the present embodiment, the loss function of use is cross entropy.Cross entropy compares a distribution and another distribution
Identical situation.Cross entropy is closer to 0, and the output of model is closer with desired output, and vice versa.
In the present embodiment, convolutional layer is made of convolution kernel, and for the feature set of one layer of neural network, convolution kernel is used
In carrying out convolutional calculation feature unit to image, general dimension is identical with target dimension.Convolution kernel size in the present embodiment
For adjustable ginseng item, the present invention is not especially limited this.The corresponding characteristic pattern of each convolution kernel, characteristic pattern is the convolution kernel
Feature Mapping.And it is weight that the parameter that machine core includes is rolled up in convolutional network.
In the present embodiment, the MobileNet framework is based on the separable convolution of depth.It is a kind of Standard convolution
Standard convolution is resolved into depth and is known as 1 × 1 convolution of the point to convolution to convolution sum by factorization convolution form.For
Single filter is applied to each input channel to convolution by MobileNets framework, depth.Then, it puts to convolution using 1 × 1
Convolution is combined into depth convolution with the output by depth to convolution.And a Standard convolution is filtered simultaneously to input and group
It closes, obtains one group of new output.The separable convolution of depth is divided into two layers, and one for the individual course of filtering and for group
The individual course of conjunction.This factorization has the function of greatly reducing calculating and model size.
In the present embodiment, after the decline of constantly gradient, the convolutional Neural net based on coding plus decoding architecture
The precision of network can be just improved, the gradient in gradient index meaning, regard each layer network as a function, penalty values can be seen
Make a function about network parameter, network parameter here includes the weight of full articulamentum, the convolution of biasing and convolutional layer
Core, biasing.If function is most worth, extreme value must be first taken, the purpose of gradient decline is to obtain each extreme point to calculate penalty values most
The value of network parameter as a child.To obtain a highest network structure of precision.
Preferably, described to pre-process to CT image to be predicted, obtaining test sample includes:
The pixel value of the CT image to be predicted is normalized into preset section, obtains test sample.
Specifically, the pixel value of the CT image to be predicted is normalized between preset section, e.g. (- 1 ,+1)
Between, it is 100 that a threshold hypothesis, which is first arranged, when pixel value is greater than 100, which is changed to 100, the picture less than -100
Element value, is changed to -100 for the pixel value, finally to all pixel values all in 100, in the present embodiment, also to training sample
It is normalized.The present invention is not especially limited this.
Preferably, the acquisition training sample；Wherein, the training sample includes original CT image and corresponding cerebral infarction
Label includes:
Obtain the original CT image collected；
Obtain the label that cerebral infarction judgement is carried out to each original CT image.
Further, the acquisition doctor includes: to the label of each original CT image progress cerebral infarction judgement
To each original CT image, when judging the original CT image to there is cerebral infarction, acquisition has cerebral infarction mark
Label and corresponding cerebral infarction region；
To each original CT image, when judging the original CT image there is no cerebral infarction, obtain without cerebral infarction
Label.
Specifically, collecting original CT image first, in all CT images being collected into 70 percent CT is schemed
As being used as training set, collect 1 15 CT image as verifying, last remaining 1 15 CT image is made
For test set, in the training stage, the initial convolutional neural networks traverse the training set and test set, original to obtain
The cerebral infarction region of CT image and the corresponding original CT image being labeled by doctor is tested in completion traversal and then use
Card collection is trained the initial convolutional neural networks, saves weight file.
Specifically, in prediction, original brain CT image is inputted, by coding, is first referring to Fig. 3
Mobilenet framework replaces general convolution operation to reduce network parameter, followed by porous space gold using the revoluble product of depth
Word tower basin module is repaired the loss of cut zone boundary information with reference to the feature of different scale, will finally be passed through
The preceding N layers of original characteristic pattern that Mobilenet framework extracts are added to be extracted by porous spatial pyramid pond module
In high-level characteristic figure out, and be input to decoded portion, 1*1Conv changed into 3*3Conv, after the completion of the training based on
The convolutional neural networks of coding plus decoding architecture carry out 2 classification (infraction or non-stalks to each pixel in original CT image
Plug), the result of each pixel classification is finally spliced into focal area segmentation figure.Finally obtain the cerebral infarction focal zone of prediction
Domain, and it is used to the coincidence of assessment prediction segmentation result (prediction focal zone) and legitimate reading (being really focal zone) by dice coefficient
Degree.
Implement the present embodiment to have the following beneficial effects:
Training sample is obtained first；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；So
The convolutional neural networks based on coding plus decoding architecture that training is completed are obtained afterwards；Wherein, it is described training complete based on coding
Add the convolutional neural networks of decoding architecture by carrying out convolutional neural networks to the training sample and initial convolutional neural networks
Training is assessed prediction result using dice coefficient when training, according to the assessment result to the initial convolutional network
Carry out backpropagation.In the training process, using the original CT image and corresponding cerebral infarction label to initial convolutional Neural
Network training, and default result and corresponding cerebral infarction label are assessed by dice coefficient, backpropagation is carried out to reach
To optimal convolutional neural networks parameter；CT image to be predicted is pre-processed, test sample is obtained；According to the training
The convolutional neural networks based on coding plus decoding architecture completed carry out cerebral infarction prediction to the CT image to be predicted, obtain
To prediction result.By it is described based on coding plus decoding architecture convolutional neural networks image recognition, can predict whether
There is the region of cerebral infarction and cerebral infarction.
Embodiment three
Referring to fig. 4, the structural schematic diagram of the AI prediction meanss for the cerebral infarction that third embodiment of the invention provides；
A kind of AI prediction meanss of cerebral infarction, comprising:
Training sample obtains module 31, for obtaining training sample；Wherein, the training sample include original CT image and
Corresponding cerebral infarction label；
Network obtains module 32, for obtaining the convolutional neural networks based on coding plus decoding architecture of training completion；Its
In, the convolutional neural networks based on coding plus decoding architecture that the training is completed pass through to the training sample and initial convolution
Neural network carries out convolutional neural networks training, is assessed using dice coefficient prediction result when training, according to institute's commentary
Estimate result and backpropagation is carried out to the initial convolutional network.
Test sample obtains module 33 and obtains test sample for pre-processing to CT image to be predicted；
Prediction module, the convolutional neural networks based on coding plus decoding architecture for being completed according to the training, to institute
It states CT image to be predicted and carries out cerebral infarction prediction, obtain prediction result.
Preferably, the network acquisition module 32 includes:
Initial network acquiring unit, for obtaining initial convolutional neural networks；
Characteristic pattern acquiring unit, for being encoded based on the initial convolutional neural networks to the original CT image,
The corresponding characteristic pattern of each layer of the original CT image is extracted by convolution operation；
Predicting unit is mentioned described in processing for being decoded based on the initial convolutional neural networks to the characteristic pattern
Characteristic pattern after taking, and cerebral infarction focal zone is predicted；
Assessment unit is assessed for prediction result, by prediction result set and the corresponding cerebral infarction tag set
Penalty values of the cross entropy as prediction, and solved by dice coefficient, to realize backpropagation.
Preferably, the characteristic pattern acquiring unit includes:
Based on Mobilenet framework and porous spatial pyramid pond module, it is right to extract each layer of the original CT image
The characteristic pattern answered.
Preferably, the predicting unit includes:
When handling the characteristic pattern after the extraction, the preceding N layers of initial corresponding characteristic pattern obtained by coding are added,
To repair the loss of cut zone boundary information；Wherein, N is greater than 1.
Preferably, the test sample acquisition module 33 includes:
The pixel value of the CT image to be predicted is normalized into preset section, obtains test sample.
Preferably, the training sample acquisition module 31 includes:
CT image acquisition unit, for obtaining the original CT image collected；
Label acquiring unit, for obtaining the label for carrying out cerebral infarction judgement to each original CT image.
Preferably, the label acquiring unit includes:
To each original CT image, when judging the original CT image to there is cerebral infarction, acquisition has cerebral infarction mark
Label and corresponding cerebral infarction region；
To each original CT image, when judging the original CT image there is no cerebral infarction, obtain without cerebral infarction
Label.
Implement the present embodiment to have the following beneficial effects:
Training sample is obtained first；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；So
The convolutional neural networks based on coding plus decoding architecture that training is completed are obtained afterwards；Wherein, it is described training complete based on coding
Add the convolutional neural networks of decoding architecture by carrying out convolutional neural networks to the training sample and initial convolutional neural networks
Training is assessed prediction result using dice coefficient when training, according to the assessment result to the initial convolutional network
Carry out backpropagation.In the training process, using the original CT image and corresponding cerebral infarction label to initial convolutional Neural
Network training, and default result and corresponding cerebral infarction label are assessed by dice coefficient, backpropagation is carried out to reach
To optimal convolutional neural networks parameter；CT image to be predicted is pre-processed, test sample is obtained；According to the training
The convolutional neural networks based on coding plus decoding architecture completed carry out cerebral infarction prediction to the CT image to be predicted, obtain
To prediction result.By it is described based on coding plus decoding architecture convolutional neural networks image recognition, can predict whether
There is the region of cerebral infarction and cerebral infarction.
It should be noted that the apparatus embodiments described above are merely exemplary, wherein described be used as separation unit
The unit of explanation may or may not be physically separated, and component shown as a unit can be or can also be with
It is not physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to actual
It needs that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.In addition, device provided by the invention
In embodiment attached drawing, the connection relationship between module indicate between them have communication connection, specifically can be implemented as one or
A plurality of communication bus or signal wire.Those of ordinary skill in the art are without creative efforts, it can understand
And implement.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
It should be noted that in the above-described embodiments, all emphasizing particularly on different fields to the description of each embodiment, in some embodiment
In the part that is not described in, reference can be made to the related descriptions of other embodiments.Secondly, those skilled in the art should also know
It knows, the embodiments described in the specification are all preferred embodiments, and related movement and simulation must be that the present invention must
Must.
Claims (10)
1. a kind of AI prediction technique of cerebral infarction characterized by comprising
Obtain training sample；Wherein, the training sample includes original CT image and corresponding cerebral infarction label；
Obtain the convolutional neural networks based on coding plus decoding architecture that training is completed；Wherein, it is described training complete based on volume
The convolutional neural networks of code plus decoding architecture are by carrying out convolutional Neural net to the training sample and initial convolutional neural networks
Network training is assessed prediction result using dice coefficient when training, according to the assessment result to the initial convolution net
Network carries out backpropagation.
CT image to be predicted is pre-processed, test sample is obtained；
According to it is described training complete based on coding plus decoding architecture convolutional neural networks, to the CT image to be predicted into
The prediction of row cerebral infarction, obtains prediction result.
2. the AI prediction technique of cerebral infarction according to claim 1, which is characterized in that it is described obtain training complete based on
The convolutional neural networks of coding plus decoding architecture；Wherein, the convolutional Neural based on coding plus decoding architecture that the training is completed
Network uses dice system when training by carrying out convolutional neural networks training to the training sample and initial convolutional neural networks
Several pairs of prediction results are assessed, and carry out backpropagation to the initial convolutional network according to the assessment result, comprising:
Obtain initial convolutional neural networks；
The original CT image is encoded based on the initial convolutional neural networks, is extracted by convolution operation described original
The corresponding characteristic pattern of each layer of CT image；
The characteristic pattern is decoded based on the initial convolutional neural networks, the characteristic pattern after handling the extraction, and right
The prediction of cerebral infarction focal zone；
Prediction result assessment, using prediction result set with the cross entropy of the corresponding cerebral infarction tag set as the damage of prediction
Mistake value, and solved by dice coefficient, to realize backpropagation.
3. the AI prediction technique of cerebral infarction according to claim 2, which is characterized in that described based on the initial convolution mind
The original CT image is encoded through network, the corresponding characteristic pattern of each layer of the original CT image is extracted by convolution operation
Include:
Based on Mobilenet framework and porous spatial pyramid pond module, it is corresponding to extract each layer of the original CT image
Characteristic pattern.
4. the AI prediction technique of cerebral infarction according to claim 2, which is characterized in that described based on the initial convolution mind
The characteristic pattern is decoded through network, the characteristic pattern after handling the extraction, and includes: to the prediction of cerebral infarction focal zone
When handling the characteristic pattern after the extraction, the preceding N layers of initial corresponding characteristic pattern obtained by coding are added, to repair
Subdivision cuts the loss of zone boundary information；Wherein, N is greater than 1.
5. the AI prediction technique of cerebral infarction according to claim 1, which is characterized in that it is described will be to CT image to be predicted
It is pre-processed, obtaining test sample includes:
The pixel value of the CT image to be predicted is normalized into preset section, obtains test sample.
6. the AI prediction technique of cerebral infarction according to claim 1, which is characterized in that the acquisition training sample；Wherein,
The training sample includes original CT image and corresponding cerebral infarction label includes:
Obtain the original CT image collected；
Obtain the label that cerebral infarction judgement is carried out to each original CT image.
7. the AI prediction technique of cerebral infarction according to claim 6, which is characterized in that the acquisition doctor is to each described
Original CT image carry out cerebral infarction judgement label include:
To each original CT image, when judging the original CT image to there is cerebral infarction, acquisition have cerebral infarction label and
Corresponding cerebral infarction region；
To each original CT image, when judging the original CT image there is no cerebral infarction, obtain without cerebral infarction label.
8. a kind of AI prediction meanss of cerebral infarction characterized by comprising
Training sample obtains module, for obtaining training sample；Wherein, the training sample includes original CT image and corresponding
Cerebral infarction label；
Network obtains module, for obtaining the convolutional neural networks based on coding plus decoding architecture of training completion；Wherein, described
The convolutional neural networks based on coding plus decoding architecture that training is completed pass through to the training sample and initial convolution nerve net
Network carries out convolutional neural networks training, is assessed using dice coefficient prediction result when training, according to the assessment result
Backpropagation is carried out to the initial convolutional network.
Test sample obtains module and obtains test sample for pre-processing to CT image to be predicted；
Prediction module, for according to it is described training complete based on coding plus decoding architecture convolutional neural networks, to it is described to
The CT image of prediction carries out cerebral infarction prediction, obtains prediction result.
9. a kind of pre- measurement equipment of the AI of cerebral infarction, which is characterized in that including processor, memory and be stored in the memory
In and be configured as the computer program executed by the processor, the processor is realized such as when executing the computer program
The AI prediction technique of cerebral infarction described in claim 1 to 7 any one.
10. a kind of computer readable storage medium, which is characterized in that the computer readable storage medium includes the calculating of storage
Machine program, wherein equipment where controlling the computer readable storage medium in computer program operation is executed as weighed
Benefit require any one of 1 to 7 described in cerebral infarction AI prediction technique.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810983952.8A CN109259784A (en) | 2018-08-27 | 2018-08-27 | AI prediction technique, device, equipment and the storage medium of cerebral infarction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810983952.8A CN109259784A (en) | 2018-08-27 | 2018-08-27 | AI prediction technique, device, equipment and the storage medium of cerebral infarction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109259784A true CN109259784A (en) | 2019-01-25 |
Family
ID=65154357
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810983952.8A Pending CN109259784A (en) | 2018-08-27 | 2018-08-27 | AI prediction technique, device, equipment and the storage medium of cerebral infarction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109259784A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110033038A (en) * | 2019-04-09 | 2019-07-19 | 电子科技大学 | The intelligent Algorithms of Selecting of aerothermodynamics experiment data |
| CN110189299A (en) * | 2019-04-22 | 2019-08-30 | 华中科技大学 | A kind of cerebrovascular events automatic identifying method and system based on MoileNet |
| CN110647810A (en) * | 2019-08-16 | 2020-01-03 | 西北大学 | Method and device for constructing and identifying radio signal image identification model |
| CN110766048A (en) * | 2019-09-18 | 2020-02-07 | 平安科技（深圳）有限公司 | Image content identification method and device, computer equipment and storage medium |
| CN110934606A (en) * | 2019-10-31 | 2020-03-31 | 上海杏脉信息科技有限公司 | Cerebral apoplexy early-stage flat-scan CT image evaluation system and method and readable storage medium |
| CN111127366A (en) * | 2019-12-26 | 2020-05-08 | 山东浪潮人工智能研究院有限公司 | Portrait picture restoration method, device and equipment |
| CN111369528A (en) * | 2020-03-03 | 2020-07-03 | 重庆理工大学 | Coronary artery angiography image stenosis region marking method based on deep convolutional network |
| WO2020168697A1 (en) * | 2019-02-22 | 2020-08-27 | 未艾医疗技术(深圳)有限公司 | Ai identification method of embolism based on vrds 4d medical image, and product |
| CN111667458A (en) * | 2020-04-30 | 2020-09-15 | 杭州深睿博联科技有限公司 | Method and device for detecting early acute cerebral infarction in flat-scan CT |
| CN112463638A (en) * | 2020-12-11 | 2021-03-09 | 清华大学深圳国际研究生院 | Fuzzy test method based on neural network and computer readable storage medium |
| CN112767350A (en) * | 2021-01-19 | 2021-05-07 | 深圳麦科田生物医疗技术股份有限公司 | Method, device, equipment and storage medium for predicting maximum interval of thromboelastogram |
| CN112991192A (en) * | 2019-12-18 | 2021-06-18 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system thereof |
| CN115187512A (en) * | 2022-06-10 | 2022-10-14 | 珠海市人民医院 | Hepatocellular carcinoma great vessel invasion risk prediction method, system, device and medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110130371A (en) * | 2010-05-27 | 2011-12-05 | 동국대학교 산학협력단 | Method of fluorescent thrombus imaging-based prediction of infarct volume in a stroke model |
| CN106021941A (en) * | 2016-05-27 | 2016-10-12 | 成都信汇聚源科技有限公司 | Method for obtaining risk predictions of sudden death with weight value matrices of artificial neural network |
| CN106250707A (en) * | 2016-08-12 | 2016-12-21 | 王双坤 | A kind of based on degree of depth learning algorithm process head construction as the method for data |
| WO2018005820A1 (en) * | 2016-06-29 | 2018-01-04 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for utlizing brain structural characteristics for predicting a diagnosis of a neurobehavioral disorder |
| CN108257674A (en) * | 2018-01-24 | 2018-07-06 | 龙马智芯（珠海横琴）科技有限公司 | Disease forecasting method and apparatus, equipment, computer readable storage medium |
| CN109223002A (en) * | 2018-08-27 | 2019-01-18 | 上海铱硙医疗科技有限公司 | Self-closing disease illness prediction technique, device, equipment and storage medium |
-
2018
- 2018-08-27 CN CN201810983952.8A patent/CN109259784A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110130371A (en) * | 2010-05-27 | 2011-12-05 | 동국대학교 산학협력단 | Method of fluorescent thrombus imaging-based prediction of infarct volume in a stroke model |
| CN106021941A (en) * | 2016-05-27 | 2016-10-12 | 成都信汇聚源科技有限公司 | Method for obtaining risk predictions of sudden death with weight value matrices of artificial neural network |
| WO2018005820A1 (en) * | 2016-06-29 | 2018-01-04 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for utlizing brain structural characteristics for predicting a diagnosis of a neurobehavioral disorder |
| CN106250707A (en) * | 2016-08-12 | 2016-12-21 | 王双坤 | A kind of based on degree of depth learning algorithm process head construction as the method for data |
| CN108257674A (en) * | 2018-01-24 | 2018-07-06 | 龙马智芯（珠海横琴）科技有限公司 | Disease forecasting method and apparatus, equipment, computer readable storage medium |
| CN109223002A (en) * | 2018-08-27 | 2019-01-18 | 上海铱硙医疗科技有限公司 | Self-closing disease illness prediction technique, device, equipment and storage medium |
Non-Patent Citations (7)
| Title |
|---|
| CHEN MIN,ET AL: "Disease Prediction by Machine Learning Over Big Data From Healthcare Communities", 《IEEE ACCESS》 * |
| SINAT_29552923: "金字塔池化过程及其优势", 《HTTPS://BLOG.CSDN.NET/SINAT_29552923/ARTICLE/DETAILS/72638795》 * |
| 亢寒 等: "基于深度学习的医学图像分割技术", 《人工智能》 * |
| 刘兴旺 等: "一种融合AutoEncoder与CNN的混合算法用于图像特征提取", 《计算机应用研究》 * |
| 姜枫 等: "基于内容的图像分割方法综述", 《软件学报》 * |
| 贾文娟 等: "基于影像的病脑检测方法综述", 《计算机工程与应用》 * |
| 郭少军 等: "应用空间金字塔池化LBP特征的舰船检测识别", 《激光与红外》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020168697A1 (en) * | 2019-02-22 | 2020-08-27 | 未艾医疗技术(深圳)有限公司 | Ai identification method of embolism based on vrds 4d medical image, and product |
| CN110033038A (en) * | 2019-04-09 | 2019-07-19 | 电子科技大学 | The intelligent Algorithms of Selecting of aerothermodynamics experiment data |
| CN110189299A (en) * | 2019-04-22 | 2019-08-30 | 华中科技大学 | A kind of cerebrovascular events automatic identifying method and system based on MoileNet |
| CN110189299B (en) * | 2019-04-22 | 2020-12-08 | 华中科技大学 | Cerebrovascular event automatic identification method and system based on MobileNet |
| CN110647810A (en) * | 2019-08-16 | 2020-01-03 | 西北大学 | Method and device for constructing and identifying radio signal image identification model |
| CN110766048A (en) * | 2019-09-18 | 2020-02-07 | 平安科技（深圳）有限公司 | Image content identification method and device, computer equipment and storage medium |
| CN110934606A (en) * | 2019-10-31 | 2020-03-31 | 上海杏脉信息科技有限公司 | Cerebral apoplexy early-stage flat-scan CT image evaluation system and method and readable storage medium |
| CN112991192A (en) * | 2019-12-18 | 2021-06-18 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system thereof |
| CN111127366A (en) * | 2019-12-26 | 2020-05-08 | 山东浪潮人工智能研究院有限公司 | Portrait picture restoration method, device and equipment |
| CN111369528A (en) * | 2020-03-03 | 2020-07-03 | 重庆理工大学 | Coronary artery angiography image stenosis region marking method based on deep convolutional network |
| CN111369528B (en) * | 2020-03-03 | 2022-09-09 | 重庆理工大学 | Coronary artery angiography image stenosis region marking method based on deep convolutional network |
| CN111667458A (en) * | 2020-04-30 | 2020-09-15 | 杭州深睿博联科技有限公司 | Method and device for detecting early acute cerebral infarction in flat-scan CT |
| CN111667458B (en) * | 2020-04-30 | 2023-09-01 | 杭州深睿博联科技有限公司 | Early acute cerebral infarction detection method and device in flat scanning CT |
| CN112463638A (en) * | 2020-12-11 | 2021-03-09 | 清华大学深圳国际研究生院 | Fuzzy test method based on neural network and computer readable storage medium |
| CN112463638B (en) * | 2020-12-11 | 2022-09-20 | 清华大学深圳国际研究生院 | Fuzzy test method based on neural network and computer readable storage medium |
| CN112767350A (en) * | 2021-01-19 | 2021-05-07 | 深圳麦科田生物医疗技术股份有限公司 | Method, device, equipment and storage medium for predicting maximum interval of thromboelastogram |
| CN112767350B (en) * | 2021-01-19 | 2024-04-26 | 深圳麦科田生物医疗技术股份有限公司 | Method, device, equipment and storage medium for predicting maximum section of thromboelastography |
| CN115187512A (en) * | 2022-06-10 | 2022-10-14 | 珠海市人民医院 | Hepatocellular carcinoma great vessel invasion risk prediction method, system, device and medium |
| CN115187512B (en) * | 2022-06-10 | 2024-01-30 | 珠海市人民医院 | Method, system, device and medium for predicting invasion risk of large blood vessel of hepatocellular carcinoma |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109259784A (en) | AI prediction technique, device, equipment and the storage medium of cerebral infarction | |
| CN109410185B (en) | A kind of image partition method, device and storage medium | |
| CN109785928A (en) | Diagnosis and treatment proposal recommending method, device and storage medium | |
| Wang et al. | Bi-directional dermoscopic feature learning and multi-scale consistent decision fusion for skin lesion segmentation | |
| CN105011911B (en) | A kind of network-based Chinese medicine health consultation system | |
| CN108898175A (en) | Area of computer aided model building method based on deep learning gastric cancer pathological section | |
| Akyeramfo-Sam et al. | A web-based skin disease diagnosis using convolutional neural networks | |
| CN107492099A (en) | Medical image analysis method, medical image analysis system and storage medium | |
| CN109919928A (en) | Detection method, device and the storage medium of medical image | |
| CN110348515A (en) | Image classification method, image classification model training method and device | |
| CN107145756A (en) | A kind of stroke types Forecasting Methodology and device | |
| CN109223002A (en) | Self-closing disease illness prediction technique, device, equipment and storage medium | |
| CN110070540A (en) | Image generating method, device, computer equipment and storage medium | |
| CN106170246A (en) | For four-dimensional (4D) stream equipment of nuclear magnetic resonance, method and product | |
| CN109346159A (en) | Case image classification method, device, computer equipment and storage medium | |
| JP2007503629A (en) | Neural network for processing a data array having existing phase information such as an image and method of using the network | |
| CN108682009A (en) | A kind of Alzheimer's disease prediction technique, device, equipment and medium | |
| CN110458829A (en) | Image quality control method, device, equipment and storage medium based on artificial intelligence | |
| CN111785366B (en) | Patient treatment scheme determination method and device and computer equipment | |
| CN107784322A (en) | Abnormal deviation data examination method, device, storage medium and program product | |
| CN109961446A (en) | CT/MR three-dimensional image segmentation processing method, device, equipment and medium | |
| CN113496494A (en) | Two-dimensional skeleton segmentation method and device based on DRR simulation data generation | |
| CN109222902A (en) | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance | |
| CN115018809A (en) | Target area segmentation and identification method and system of CT image | |
| CN112001325A (en) | Prediction method and device for unsafe power distribution behaviors and server |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190125 |