CN108399619A - The system and device of medical diagnosis - Google Patents
The system and device of medical diagnosis Download PDFInfo
- Publication number
- CN108399619A CN108399619A CN201810171115.5A CN201810171115A CN108399619A CN 108399619 A CN108399619 A CN 108399619A CN 201810171115 A CN201810171115 A CN 201810171115A CN 108399619 A CN108399619 A CN 108399619A
- Authority
- CN
- China
- Prior art keywords
- medical image
- network
- neural network
- nervus opticus
- data
- 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.)
- Granted
Links
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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
Present disclose provides a kind of systems of medical diagnosis.The system comprises one or more processors and memories.The memory is stored with computer-readable instruction.Wherein, the processor is made to realize when described instruction is executed by processor：At least one medical image is obtained, at least one medical image is the medical image that single pass obtains；And output has the diagnosis report data of mapping relations at least one medical image.The disclosure additionally provides a kind of device of medical diagnosis, a kind of training method carrying out medical diagnosis using robot.
Description
Technical field
This disclosure relates to the system and device of a kind of medical diagnosis.
Background technology
Artificial intelligence is the main trend of future development.The development of current manual's intelligence has caused the extensive of all trades and professions
Attention.Image recognition technology has been widely applied at present.Constantly enhance with the generalization ability of artificial intelligence, utilizes
Artificial intelligence carries out image recognition to replace previous manpower to identify that image is not only efficient but also more accurate.In medical department
In system, dependence medical practitioner is usually required by medical imaging acquisition diagnosis report, medical image is understood.It is this artificial
Understanding medical image can take considerable time to obtain the mode of diagnosis report, and can consume the great effort of doctor.
Invention content
The first aspect of the disclosure provides a kind of method of medical diagnosis, including：Obtain at least one medical image, institute
It is the medical image that single pass obtains to state at least one medical image；Output has mapping at least one medical image
The diagnosis report data of relationship.
Optionally, it exports and includes with diagnosis report data of at least one medical image with mapping relations：By institute
It states at least one medical image and is input to neural network；And obtain the output of the neural network, wherein the neural network
Output include the diagnosis report data.
Optionally, the neural network includes at least one first nerves network and at least one nervus opticus network；It will
At least one medical image is input to neural network, including：At least one medical image is input to the first nerves
Network is to extract at least one characteristic of at least one medical image；And it is at least one characteristic is defeated
Enter to the nervus opticus network, there is mapping at least one medical image to be obtained by the nervus opticus network
The diagnosis report data of relationship；Wherein, the first nerves network and nervus opticus network are same type or difference
Type.
Optionally, the first nerves network includes convolutional neural networks；And the nervus opticus network includes cycle
Neural network.
Optionally, when it is described at least a characteristic includes multiple characteristics when, by least one characteristic
It is input to the nervus opticus network, is reflected with obtaining to have at least one medical image by the nervus opticus network
The diagnosis report data of relationship are penetrated, including：It is a compressive features data by the multiple feature data compression；It will be described
Compressive features data are input to the nervus opticus network, to be obtained and at least one doctor by the nervus opticus network
Treating image has the diagnosis report data of mapping relations.
Optionally, it is a compressive features data by the multiple feature data compression, including：According to the multiple feature
The correlation of each characteristic content status data current with the nervus opticus network in data determines described more
The compression weight of each characteristic in a characteristic, wherein correlation more high compression weight are bigger；According to the pressure
Contracting weight is weighted averagely, to obtain the compressive features data the multiple characteristic.
The second aspect of the disclosure provides a kind of device of medical diagnosis, including：Medical image acquisition module, is used for
At least one medical image is obtained, at least one medical image is the medical image that single pass obtains；And diagnosis report
Data outputting module is accused, for exporting the diagnosis report data that there are mapping relations at least one medical image.
Optionally, the diagnosis report data outputting module includes：Neural network input submodule is used for by described at least
One medical image is input to neural network；And neural network exports acquisition submodule, for obtaining the neural network
Output, wherein the output of the neural network includes the diagnosis report data.
Optionally, the neural network includes at least one first nerves network and at least one nervus opticus network, institute
Stating neural network input submodule includes：First nerves network inputs unit, for inputting at least one medical image
To first nerves network to extract at least one characteristic of at least one medical image；And nervus opticus network is defeated
Enter unit, at least one characteristic to be input to the nervus opticus network, to pass through the nervus opticus net
Network obtains the diagnosis report data for having mapping relations at least one medical image；Wherein：The first nerves
Network and nervus opticus network are same type or different type.
Optionally, it is described at least a characteristic includes multiple characteristics when, by least one characteristic
It is input to the nervus opticus network, is reflected with obtaining to have at least one medical image by the nervus opticus network
The diagnosis report data of relationship are penetrated, including：It is a compressive features data by the multiple feature data compression；And it will
The compressive features data are input to the nervus opticus network, to be obtained and described at least one by the nervus opticus network
A medical image has the diagnosis report data of mapping relations.
Optionally, it is a compressive features data by the multiple feature data compression, including：According to the multiple feature
The correlation of each characteristic content status data current with the nervus opticus network in data determines described more
The compression weight of each characteristic in a characteristic, wherein correlation more high compression weight are bigger；According to the pressure
Contracting weight is weighted averagely, to obtain the compressive features data the multiple characteristic.
The third aspect of the disclosure provides a kind of system of medical diagnosis, including one or more processors, Yi Jicun
Reservoir.The memory is stored with computer-readable instruction.Described instruction makes the processor when being executed by the processor
Realize the method described in the first aspect according to the disclosure.
The fourth aspect of the disclosure provides a kind of non-volatile memory medium, is stored with computer executable instructions, institute
Instruction is stated when executed for realizing the method described in the first aspect according to the disclosure.
5th aspect of the disclosure provides a kind of computer program, and the computer program, which includes that computer is executable, to be referred to
It enables, described instruction is when executed for realizing the method described in the first aspect according to the disclosure.
6th aspect of the disclosure provides a kind of training method carrying out medical diagnosis using robot, including：It is near
A few medical image is input to neural network to obtain the output of the neural network, wherein at least one medical treatment figure
As the medical image obtained for single pass, the output of the neural network includes for describing at least one medical image
Character data；When the consistency of output and the model answer of the neural network does not meet preset condition, repeat
The input operation, has been trained when the output of the neural network meets preset condition with the model answer consistency
At wherein the model answer includes the diagnosis report data for having mapping relations at least one medical image；Output
The neural network that training is completed.
7th aspect of the disclosure provides a kind of system for the training carrying out medical diagnosis using robot, including one
Or multiple processors and memory.The memory is stored with computer-readable instruction.Described instruction is held by the processor
Make the method described in the 6th aspect of the processor realization according to the disclosure when row.
The eighth aspect of the disclosure provides a kind of non-volatile memory medium, is stored with computer executable instructions, institute
Instruction is stated when executed for realizing the method described in the 6th aspect according to the disclosure.
9th aspect of the disclosure provides a kind of computer program, and the computer program, which includes that computer is executable, to be referred to
It enables, described instruction is when executed for realizing the method described in the 6th aspect according to the disclosure.
Tenth aspect of the disclosure provides a kind of processing method of medical image.It is described to include：Obtain at least one doctor
Image is treated, at least one medical image is the medical image that single pass obtains；And it exports and at least one doctor
Treating image has the descriptive character data of mapping relations；The wherein described descriptive character data includes by natural language to institute
State the data that at least one medical image itself is described.
Optionally, it exports and includes with descriptive character data of at least one medical image with mapping relations：It will
At least one medical image is input to neural network；And obtain the output of the neural network, wherein the nerve net
The output of network includes the descriptive character data.
Optionally, the neural network includes at least one first nerves network and at least one nervus opticus network；It will
At least one medical image is input to neural network, including：At least one medical image is input to first nerves network
To extract at least one characteristic of at least one medical image；And at least one characteristic is input to
The nervus opticus network has mapping relations to be obtained by the nervus opticus network at least one medical image
The descriptive character data；Wherein, the first nerves network and nervus opticus network are same type or inhomogeneity
Type.
Optionally, the first nerves network includes convolutional neural networks；And the nervus opticus network includes cycle
Neural network.
Optionally, when it is described at least a characteristic includes multiple characteristics when, by least one characteristic
It is input to the nervus opticus network, is reflected with obtaining to have at least one medical image by the nervus opticus network
The descriptive character data of relationship is penetrated, including：It is a compressive features data by the multiple feature data compression；By institute
State compressive features data and be input to the nervus opticus network, with by the nervus opticus network obtain with it is described at least one
Medical image has the descriptive character data of mapping relations.
Optionally, it is a compressive features data by the multiple feature data compression, including：According to the multiple feature
The correlation of each characteristic content status data current with the nervus opticus network in data determines described more
The compression weight of each characteristic in a characteristic, wherein correlation more high compression weight are bigger；According to the pressure
Contracting weight is weighted averagely, to obtain the compressive features data the multiple characteristic.
Tenth one side of the disclosure provides a kind of processing unit of medical image, including：Medical image acquisition module,
For obtaining at least one medical image, at least one medical image is the medical image that single pass obtains；And word
Data outputting module is accorded with, for exporting the descriptive character data that there are mapping relations at least one medical image；Its
Described in descriptive character data include the data that described at least one medical image itself is described by natural language.
12nd aspect of the disclosure provides a kind of processing system of medical image, including one or more processors,
And memory.The memory is stored with computer-readable instruction.Described instruction makes described when being executed by the processor
Processor realizes the processing method of the medical image described in the tenth aspect according to the disclosure.
13rd aspect of the disclosure provides a kind of non-volatile memory medium, is stored with computer executable instructions,
Described instruction is when executed for realizing the processing method of the medical image described in the tenth aspect according to the disclosure.
The fourteenth aspect of the disclosure provides a kind of computer program, and the computer program includes that computer is executable
Instruction, described instruction is when executed for realizing the processing method of the medical image described in the tenth aspect according to the disclosure.
15th aspect of the disclosure provides a kind of training method handling medical image using robot, including：It will
At least one medical image is input to neural network to obtain the output of the neural network, wherein at least one medical treatment
Image is the medical image that single pass obtains, and the output of the neural network includes for describing at least one medical treatment figure
The character data of picture；When the consistency of output and the model answer of the neural network does not meet preset condition, repetition is held
The row input operation, has been trained when the output of the neural network meets preset condition with the model answer consistency
At wherein the model answer includes the predetermined descriptive word at least one medical image with mapping relations
Data are accorded with, the descriptive character data includes that described at least one medical image itself is described by natural language
Data；And the neural network that output training is completed.
16th aspect of the disclosure provides a kind of training system handling medical image using robot, including one
Or multiple processors and memory.The memory is stored with computer-readable instruction.Described instruction is held by the processor
Make the instruction that medical image is handled using robot described in the 15th aspect of the processor realization according to the disclosure when row
Practice method.
17th aspect of the disclosure provides a kind of non-volatile memory medium, is stored with computer executable instructions,
Described instruction handles medical image for realizing described in the 15th aspect according to the disclosure using robot when executed
Training method.
18th aspect of the disclosure provides a kind of computer program, and the computer program includes that computer is executable
Instruction, described instruction handle medical treatment for realizing described in the 15th aspect according to the disclosure using robot when executed
The training method of image.
Description of the drawings
In order to which the disclosure and its advantage is more fully understood, referring now to being described below in conjunction with attached drawing, wherein：
Figure 1A diagrammatically illustrates the flow chart of the medical diagnostic method according to the embodiment of the present disclosure；
Figure 1B diagrammatically illustrates the flow chart of the processing method of the medical image according to the embodiment of the present disclosure；
Fig. 2A diagrammatically illustrates the flow chart of the medical diagnostic method according to another embodiment of the disclosure；
Fig. 2 B diagrammatically illustrate the flow chart of the processing method of the medical image according to another embodiment of the disclosure；
Fig. 3 A are diagrammatically illustrated in Fig. 2A according to another embodiment of the disclosure and are input at least one medical image
The flow chart of the method for neural network；
Fig. 3 B are diagrammatically illustrated in Fig. 2 B according to another embodiment of the disclosure and are input at least one medical image
The flow chart of the method for neural network；
Fig. 4 diagrammatically illustrates the reality of the processing method of the medical diagnostic method or medical image according to the embodiment of the present disclosure
Existing scene figure；
Fig. 5 diagrammatically illustrates the block diagram of the medical diagnosis device according to the embodiment of the present disclosure；
Fig. 6 diagrammatically illustrates diagnosis report data outputting module in the medical diagnosis device according to the embodiment of the present disclosure
Block diagram；
Fig. 7 diagrammatically illustrates neural network in the medical diagnosis device according to the embodiment of the present disclosure and exports acquisition submodule
Block diagram；
Fig. 8 diagrammatically illustrates the stream of the training method that medical diagnosis is carried out using robot according to the embodiment of the present disclosure
Cheng Tu；And
Fig. 9 diagrammatically illustrates the block diagram of the medical diagnosis system according to the embodiment of the present disclosure；
Figure 10 diagrammatically illustrates the block diagram of the processing unit of the medical image according to the embodiment of the present disclosure；
Figure 11 diagrammatically illustrates the training method for handling medical image using robot according to the embodiment of the present disclosure
Flow chart；And
Figure 12 diagrammatically illustrates the block diagram of the processing system of the medical image according to the embodiment of the present disclosure.
Specific implementation mode
Hereinafter, will be described with reference to the accompanying drawings embodiment of the disclosure.However, it should be understood that these descriptions are only exemplary
, and it is not intended to limit the scope of the present disclosure.In addition, in the following description, descriptions of well-known structures and technologies are omitted, with
Avoid unnecessarily obscuring the concept of the disclosure.
Term as used herein is not intended to limit the disclosure just for the sake of description specific embodiment.It uses herein
The terms "include", "comprise" etc. show the presence of the feature, step, operation and/or component, but it is not excluded that in the presence of
Or other one or more features of addition, step, operation or component.
There are all terms (including technical and scientific term) as used herein those skilled in the art to be generally understood
Meaning, unless otherwise defined.It should be noted that term used herein should be interpreted that with consistent with the context of this specification
Meaning, without should by idealization or it is excessively mechanical in a manner of explain.
It, in general should be according to this using " in A, B and C etc. at least one " such statement is similar to
Field technology personnel are generally understood the meaning of the statement to make an explanation (for example, " with system at least one in A, B and C "
Should include but not limited to individually with A, individually with B, individually with C, with A and B, with A and C, with B and C, and/or
System etc. with A, B, C).Using " in A, B or C etc. at least one " such statement is similar to, it is general come
Say be generally understood the meaning of the statement to make an explanation (for example, " having in A, B or C at least according to those skilled in the art
One system " should include but not limited to individually with A, individually with B, individually with C, with A and B, with A and C, have
B and C, and/or system etc. with A, B, C).It should also be understood by those skilled in the art that substantially arbitrarily indicating two or more
The adversative conjunction and/or phrase of optional project shall be construed as either in specification, claims or attached drawing
It gives including one of these projects, the possibility of these projects either one or two projects.For example, phrase " A or B " should
It is understood to include the possibility of " A " or " B " or " A and B ".
Shown in the drawings of some block diagrams and/or flow chart.It should be understood that some sides in block diagram and/or flow chart
Frame or combinations thereof can be realized by computer program instructions.These computer program instructions can be supplied to all-purpose computer,
The processor of special purpose computer or other programmable data processing units, to which these instructions can be with when being executed by the processor
Create the device for realizing function/operation illustrated in these block diagrams and/or flow chart.
Therefore, the technology of the disclosure can be realized in the form of hardware and/or software (including firmware, microcode etc.).Separately
Outside, the technology of the disclosure can take the form of the computer program product on the computer-readable medium for being stored with instruction, should
Computer program product uses for instruction execution system or instruction execution system is combined to use.In the context of the disclosure
In, computer-readable medium can be the arbitrary medium can include, store, transmitting, propagating or transmitting instruction.For example, calculating
Machine readable medium can include but is not limited to electricity, magnetic, optical, electromagnetic, infrared or semiconductor system, device, device or propagation medium.
The specific example of computer-readable medium includes：Magnetic memory apparatus, such as tape or hard disk (HDD)；Light storage device, such as CD
(CD-ROM)；Memory, such as random access memory (RAM) or flash memory；And/or wire/wireless communication link.
In medical system, by medical image obtain diagnosis report usually require rely on medical practitioner to medical image into
Row is understood.If can will be in the identification of artificial intelligence application to medical image, it is possible to realize and rapidly be schemed according to medical treatment
It is not only efficient in this way as obtaining diagnosis report, but also the energy of doctor can be freed, so that doctor is more paid close attention to more
In the medical procedure for needing creative work.
Embodiment of the disclosure provides a kind of method and apparatus of medical diagnosis.This method includes obtaining at least one doctor
Image is treated, which is the medical image that single pass obtains, and exports and scheme at least one medical treatment
As having the diagnosis report data of mapping relations.
According to the medical diagnostic method of the embodiment of the present disclosure, device and system, the medical treatment that can be obtained according to single pass
Image can export corresponding diagnosis report data automatically.Thus, it is possible to which the medical treatment of traditional dependence doctor's human interpretation is schemed
The mode of picture is changed into obtains diagnosis report automatically according to medical image, to free the energy of doctor, makes doctor can
To focus more in more medical procedures for needing creative work, provided a great convenience for curative activity.
Embodiment of the disclosure additionally provides a kind of processing method of medical image and corresponding device.This method includes obtaining
It is the medical image that single pass obtains, and output and institute to take at least one medical image, at least one medical image
Stating at least one medical image has the descriptive character data of mapping relations.The wherein described descriptive character data includes passing through
The data that described at least one medical image itself is described in natural language.In accordance with an embodiment of the present disclosure, it can will cure
It treats image and is converted to corresponding descriptive character data, which can carry out nature language to the medical image
The data information understood is sayed, to provide another reading medical image approach.
Figure 1A diagrammatically illustrates the flow chart of the medical diagnostic method according to the embodiment of the present disclosure.
As shown in Figure 1A, this method includes operation S101 and S102A.
In operation S101, at least one medical image is obtained, which is the doctor that single pass obtains
Treat image.
In operation S102A, output has the diagnosis report data of mapping relations at least one medical image.
In accordance with an embodiment of the present disclosure, it can automatically be exported according to the medical image that single pass obtains and be examined accordingly
Disconnected data reporting.Schemed according to medical treatment thus, it is possible to be changed into the mode of the medical image of traditional dependence doctor's human interpretation
Diagnosis report is obtained as automatic, to free the energy of doctor, so that doctor is focused more on and more needs to create
Property labour medical procedure in, provided a great convenience for curative activity.
Figure 1B diagrammatically illustrates the flow chart of the processing method of the medical image according to the embodiment of the present disclosure.
As shown in Figure 1B, include operation S101 and operation according to the processing method of the medical image of the embodiment of the present disclosure
S102B.Wherein operation S101 can be with the associated description in reference chart 1A.
In operation S102B, output has the descriptive character data of mapping relations at least one medical image, wherein
The descriptive character data includes the data that at least one medical image itself is described by natural language.Such as it is right
In the medical image that one includes a human body part, in accordance with an embodiment of the present disclosure, which for example can be with
Be by structural form data (for example, thickness or shape etc.) of the human body parts in the natural language description medical image, by
The skeleton density distribution situations of the human body parts that the features such as the gray scale of image and/or brightness reflect, muscle distribution situation etc..
The descriptive character data can also be the prompt in the region to being paid particular attention in the medical image, such as prompt some districts
There is mutation and/or size, the position etc. in these regions in the characteristics of image in domain.In accordance with an embodiment of the present disclosure, the description
Property character data is the description to image information itself, and the diagnosis of disease cannot be immediately arrived at according to the descriptive character data
Or health status as a result.
In accordance with an embodiment of the present disclosure, by the way that at least one medical image conversion to be had to the descriptive word of mapping relations
Data are accorded with, can be the information that can be read by natural language by the information corresponding conversion at least one medical image,
To provide a kind of approach for reading at least one medical image.On this basis, medical worker can be retouched based on this
The information of at least one medical image that the property stated character data is reflected itself, and/or it is directly viewable this extremely in conjunction with its people
The information that a few medical image obtains, obtains corresponding medical diagnosis result.In this way, in application scenes, Ke Yibang
It helps medical worker etc. more accurately to grasp the information that at least one medical image itself is reflected, such as will can directly read
The information that at least one medical image is obtained is read to be mutually authenticated with the descriptive character data.Simultaneously as machine
The image recognition resolution ratio and recognition speed more much higher than human eye may be implemented, therefore the solution to medical image can be significantly improved
Read precision and efficiency.Moreover, medical image information is exported with descriptive character data, it can may be to certain to avoid manual reading of drawings
The omission of fine image information.It is real in other application scenarios, such as curative activity trainee or rigid participation medical treatment
The worker etc. trampled can obtain the information that oneself direct reading medical image obtains with according to the method for the embodiment of the present disclosure
Descriptive matter in which there information carry out mutual checksum validation, promote the promotion etc. of its medical practice level.
Fig. 2A diagrammatically illustrates the flow chart of the medical diagnostic method according to another embodiment of the disclosure.
As shown in Figure 2 A, according in the medical diagnostic method of another embodiment of the disclosure, operation S102 includes operation S201A
With operation S202A.
The medical diagnostic method of another embodiment of the specific disclosure includes operation S101, operation S201A and operation
S202A。
In operation S201A, which is input to neural network.
In operation S202A, the output of the neural network is obtained, wherein the output of the neural network includes the diagnosis report
Data.
In accordance with an embodiment of the present disclosure, diagnosis report is quickly obtained according to medical image using neural network, so as to
With effectively replace in the prior art understand medical image must rely on medical practitioner scheme, improve the efficiency of medical diagnosis.
Fig. 2 B diagrammatically illustrate the flow chart of the processing method of the medical image according to another embodiment of the disclosure.
As shown in Figure 2 B, include operation S101, operation according to the processing method of the medical image of another embodiment of the disclosure
S201B and operation S202B.
In operation S201B, which is input to neural network.
In operation S202B, the output of the neural network is obtained, wherein the output of the neural network includes the descriptive word
Accord with data.
In accordance with an embodiment of the present disclosure, the processing method of the medical image can be converted medical image by neural network
By the data of the natural language description medical image, to provide another approach for reading medical image.
Fig. 3 A, which are diagrammatically illustrated, operates S201A by least one medical treatment figure in Fig. 2A according to another embodiment of the disclosure
As the flow chart for the method for being input to neural network
As shown in Figure 3A, according to the medical diagnostic method of another embodiment of the disclosure, which includes at least one
First nerves network and at least one nervus opticus network, the first nerves network and nervus opticus network be same type or
Person's different type, operation S201A may include operation S211A and operation S212A.
In operation S211A, which is input to first nerves network to extract at least one doctor
Treat at least one characteristic of image.
For example, it may be extract the characteristic of wherein each characteristic respectively at least one medical image,
The some or all of correspondence that can be corresponded in multiple medical images extracts a characteristic.One characteristic example
It such as can be the feature vector of medical image.
In operation S212A, which is input to the nervus opticus network, to pass through the nervus opticus
Network obtains the diagnosis report data for having mapping relations at least one medical image.
Fig. 3 B, which are diagrammatically illustrated, operates S201B by least one medical treatment figure in Fig. 2 B according to another embodiment of the disclosure
As the flow chart for the method for being input to neural network.
As shown in Figure 3B, according in the processing method of the medical image of another embodiment of the disclosure, which includes
At least one first nerves network and at least one nervus opticus network, the first nerves network and nervus opticus network are same
Type or different type.Operation S201B may include operation S211B and operation S212B.
In operation S211B, which is input to first nerves network to extract at least one doctor
Treat at least one characteristic of image.It similarly, can be to operate the associated description of S211A in reference chart 3A.
In operation S212B, which is input to the nervus opticus network, to pass through described second
Neural network obtains the descriptive character data for having mapping relations at least one medical image.
In accordance with an embodiment of the present disclosure, which can be the neural network for being good at image procossing, this second
Neural network can be good at the neural network of natural language processing.In accordance with an embodiment of the present disclosure, the first nerves network packet
It includes Recognition with Recurrent Neural Network to include convolutional neural networks and the nervus opticus network.
Specifically, the function ratio of convolutional neural networks (Convolutional Neural Network, CNN) processing image
More powerful, (Recurrent Neural Network are widely used in the processing of natural language Recognition with Recurrent Neural Network RNN.
In accordance with an embodiment of the present disclosure, the first nerves network and the nervus opticus network can also be same type of god
Through network.Such as all it is convolutional neural networks or is all Recognition with Recurrent Neural Network or is all other neural networks etc..
In accordance with an embodiment of the present disclosure, when at least a characteristic includes multiple characteristics for this, this is at least one
Characteristic is input to the nervus opticus network, to have at least one medical image by nervus opticus network acquisition
Diagnosis report data or the descriptive character data of mapping relations, can for example be by multiple feature data compression
Then the compressive features data are input to the nervus opticus network by one compressive features data, to pass through the nervus opticus net
Network obtains the diagnosis report data or the descriptive character data for having mapping relations at least one medical image.
In accordance with an embodiment of the present disclosure, it is a compressive features data by multiple feature data compression, can is for example
It is related to content status data that the nervus opticus network is current according to each characteristic in multiple characteristic
Property determines the compression weight of each characteristic in multiple characteristic, and wherein correlation more high compression weight is bigger,
Then multiple characteristic is weighted averagely, to obtain the compressive features data according to the compression weight.
Below with reference to Fig. 4, in conjunction with specific embodiments to medical diagnostic method, Yi Jitu shown in Figure 1A, Fig. 2A and Fig. 3 A
The processing method of medical image shown in 1B, Fig. 2 B and Fig. 3 B is described further.
Fig. 4 diagrammatically illustrates the reality of the processing method of the medical diagnostic method or medical image according to the embodiment of the present disclosure
Existing scene figure.
As shown in figure 4, obtaining medical treatment by neural network in the medical diagnostic method or the processing method of the medical image
Image, and final output has the diagnosis report data of mapping relations or descriptive character data with the medical image.
Specifically, the characteristic of at least one medical image is extracted by first nerves network first.For example, in Fig. 4
Signal in, the characteristic of each medical image is extracted respectively by corresponding first nerves network.And one in the signal of Fig. 4
Secondary scanning process obtains multiple medical images, to which first nerves network extracts to have obtained multiple characteristics.
Then it is input to nervus opticus network after handling multiple characteristic, makes nervus opticus network according to defeated
Entering content matching and the medical image, there is the diagnostic data of mapping relations to report；Alternatively, in the doctor according to the embodiment of the present disclosure
It treats in image processing method, makes nervus opticus network that there is according to input content matching with the medical image description of mapping relations
Property character data.
Multiple characteristics are carried out processing can be being for example a compressive features number by multiple feature data compression
According to, and the compressive features data are input to the nervus opticus network.
In some embodiments, can be that multiple characteristic is weighted one characteristic of average boil down to.
In further embodiments, such as in the example of Fig. 4, half attention mechanism can be introduced, according in multiple characteristic
The correlation of each characteristic content status data current with the nervus opticus network determines in multiple characteristic
Each characteristic compression weight.
Specifically, in the example of fig. 4, first nerves network includes full convolutional neural networks (Fully
Convolutional Neural Network, FCN) and convolutional neural networks CNN.Wherein, FCN networks are the one of CNN networks
Kind, there is more powerful image-capable.
Nervus opticus network may include shot and long term memory LSTM (Long Short Term) neural network.Wherein, LSTM
Neural network belongs to a kind of special RNN neural networks, can update its internal state according to last round of input, output.
For example, it is assumed that can indicate the internal state data of LSTM neural networks with a matrix, with feature to
Amount or matrix come indicate FCN or CNN extraction medical image characteristic.It in some embodiments, can be to each
The internal state data (such as matrix) of characteristic and LSTM neural networks carries out correlation analysis, and (correlation analysis can have
It is a variety of, including Linear correlative analysis and nonlinear correlation analysis).Such as it carries out that when Linear correlative analysis two data can be obtained
Between linearly dependent coefficient.In linear correlation analysis, related coefficient is in [0,1] range, and related coefficient is closer to 1 correlation
Property is bigger, and related coefficient is closer to 0, and correlation is with regard to smaller.In accordance with an embodiment of the present disclosure, correlation is higher, according to more
Compression weight can be bigger when a characteristic obtains compressive features data.
Fig. 5 diagrammatically illustrates the block diagram of the medical diagnosis device according to the embodiment of the present disclosure.
As shown in figure 5, the device 500 includes medical image acquisition module 510 and diagnosis report data outputting module 520.
The device 500 can execute the medical diagnostic method with reference to described in figure 1A, Fig. 2A, Fig. 3 A and Fig. 4.
Specifically, medical image acquisition module 510 is for obtaining at least one medical image, at least one medical image
The medical image obtained for single pass.
Diagnosis report data outputting module 520 is used to export the diagnosis for having mapping relations at least one medical image
Data reporting.
Fig. 6 diagrammatically illustrates the frame of diagnosis report data outputting module 520 in the device 500 according to the embodiment of the present disclosure
Figure.
The diagnosis report data outputting module 520 includes that neural network input submodule 521 and neural network output obtain
Submodule 522.
Neural network input submodule 521 is used at least one medical image being input to neural network.
Neural network output acquisition submodule 522 is used to obtain the output of the neural network, wherein the neural network it is defeated
Go out including the diagnosis report data.
Fig. 7 diagrammatically illustrates neural network output acquisition submodule 522 in the device 500 according to the embodiment of the present disclosure
Block diagram.
As shown in fig. 7, in accordance with an embodiment of the present disclosure, which includes at least one first nerves network and at least
One nervus opticus network, the first nerves network and nervus opticus network are same type or different type.Neural network
It may include first nerves network inputs unit 5221 and nervus opticus network inputs unit 5222 to export acquisition submodule 522.
First nerves network inputs unit 5221 be used for by least one medical image be input to first nerves network with
Extract at least one characteristic of at least one medical image.
Nervus opticus network inputs unit 5222 is used at least one characteristic being input to the nervus opticus network,
To obtain the diagnosis report data that there are mapping relations at least one medical image by the nervus opticus network.
In accordance with an embodiment of the present disclosure, which can be according to the medical image that single pass obtains certainly
The dynamic corresponding diagnosis report data of output.Thus, it is possible to which the mode of the medical image of traditional dependence doctor's human interpretation is turned
Become to obtain diagnosis report automatically according to medical image, to free the energy of doctor, doctor is allow to focus more on
In more medical procedures for needing creative work, provided a great convenience for curative activity.
It is understood that medical image acquisition module 510, diagnosis report data outputting module 520, neural network input
Submodule 521, neural network output acquisition submodule 522, first nerves network inputs unit 5221 and nervus opticus network are defeated
Enter unit 5222 and may be incorporated in a module to realize or any one module therein can be split into multiple moulds
Block, either the multiple submodule wherein included by any one module can split or merge realization.Alternatively, these modules with
And at least partly function of one or more of submodule module can be combined at least partly function of other modules, and
It is realized in a module.According to an embodiment of the invention, medical image acquisition module 510, diagnosis report data outputting module
520, neural network input submodule 521, neural network output acquisition submodule 522, first nerves network inputs unit 5221
It can be at least implemented partly as hardware circuit with each submodule in nervus opticus network inputs unit 5222, such as existing
Field programmable gate array (FPGA), programmable logic array (PLA), system on chip, the system on substrate, the system in encapsulation,
Application-specific integrated circuit (ASIC), or can be to carry out the hardware such as any other rational method that is integrated or encapsulating to circuit or consolidate
Part realizes, or is realized with software, the appropriately combined of hardware and firmware three kinds of realization methods.Alternatively, medical image obtains
Module 510, diagnosis report data outputting module 520, neural network input submodule 521, neural network export acquisition submodule
522, each submodule included by first nerves network inputs unit 5221 and nervus opticus network inputs unit 5222 is in the block
It is at least one to be at least implemented partly as computer program module, when the program is run by computer, can execute
The function of corresponding module.
Fig. 8 diagrammatically illustrates the stream of the training method that medical diagnosis is carried out using robot according to the embodiment of the present disclosure
Cheng Tu.
As shown in figure 8, in accordance with an embodiment of the present disclosure, the training method that medical diagnosis is carried out using robot includes operation
S801~operation S803.
In operation S801, at least one medical image is input to neural network to obtain the output of the neural network,
In, which is the medical image that single pass obtains, and the output of the neural network includes for describing this
The character data of at least one medical image.
In operation S802, when the consistency of output and the model answer of the neural network does not meet preset condition, weight
Input operation is executed again, has been trained when the output of the neural network meets preset condition with the model answer consistency
At the wherein model answer includes the diagnosis report data for having mapping relations at least one medical image.
In operation S803, the neural network that training is completed is exported.
In accordance with an embodiment of the present disclosure, training neural network exports diagnosis report according to medical image.In this way, training is completed
Neural network can be quickly obtained diagnosis report according to medical image, to a certain extent can ground substitution it is existing
Medical image is understood in technology must rely on the scheme of medical practitioner, improve the efficiency of medical diagnosis.
Fig. 9 diagrammatically illustrates the block diagram of the medical diagnosis system according to the embodiment of the present disclosure.
As shown in figure 9, the medical diagnosis system 900 includes processor 910 and computer readable storage medium 920.One
In a little embodiments, which can execute the doctor above with reference to described in Figure 1A, Fig. 2A, Fig. 3 A and Fig. 4
Diagnostic method is treated, medical diagnosis is carried out according to medical image to realize.In further embodiments, which also may be used
For executing the method described above with reference to Fig. 8, carry out carrying out medical diagnosis using robot with image training robot.
Specifically, processor 910 for example may include general purpose microprocessor, instruction set processor and/or related chip group
And/or special microprocessor (for example, application-specific integrated circuit (ASIC)), etc..Processor 910 can also include being used for caching
The onboard storage device on way.Processor 910 can be performed for the medical treatment with reference to described in figure 1A, Fig. 2A, Fig. 3 A and Fig. 4
Single treatment unit either multiple processing units of the different actions of diagnostic method flow.
Computer readable storage medium 920, such as can include, store, transmitting, propagating or transmitting appointing for instruction
Meaning medium.For example, readable storage medium storing program for executing can include but is not limited to electricity, magnetic, optical, electromagnetic, infrared or semiconductor system, device,
Device or propagation medium.The specific example of readable storage medium storing program for executing includes：Magnetic memory apparatus, such as tape or hard disk (HDD)；Optical storage
Device, such as CD (CD-ROM)；Memory, such as random access memory (RAM) or flash memory；And/or wire/wireless communication chain
Road.
Computer readable storage medium 920 may include computer program 921, which may include generation
Code/computer executable instructions, make when being executed by processor 910 processor 910 execute for example above in conjunction with figure Figure 1A,
Medical diagnostic method flow described in Fig. 2A, Fig. 3 A and Fig. 4 and its any deformation, such as above in conjunction with Fig. 8 institutes
The method flow of description and its any deformation.
Computer program 921 can be configured with such as computer program code including computer program module.Example
Such as, in the exemplary embodiment, the code in computer program 921 may include one or more program modules, such as including
921A, module 921B ....It should be noted that the dividing mode and number of module are not fixed, those skilled in the art can
To be combined using suitable program module or program module according to actual conditions, when these program modules are combined by processor 910
When execution so that processor 910, which can execute, for example examines above in conjunction with the medical treatment described in Figure 1A, Fig. 2A, Fig. 3 A and Fig. 4
Disconnected method flow and its any deformation, such as above in conjunction with method flow described in Fig. 8 and its any deformation.
According to an embodiment of the invention, medical image acquisition module 510, diagnosis report data outputting module 520, nerve net
Network input submodule 521, neural network output acquisition submodule 522, first nerves network inputs unit 5221 and nervus opticus
At least one computer program that can be implemented as describing with reference to figure 9 of each submodule included by network inputs unit 5222
Corresponding operating described above may be implemented when being executed by processor 910 in module.
Figure 10 diagrammatically illustrates the block diagram of the processing unit 1000 according to the medical image of the embodiment of the present disclosure.
As shown in Figure 10, the processing unit 1000 of the medical image includes medical image acquisition module 1010 and character data
Output module 1020.The processing unit 1000 can execute the medical treatment figure with reference to described in figure 1B, Fig. 2 B, Fig. 3 B and Fig. 4
The processing method of picture.
For medical image acquisition module 1010 for obtaining at least one medical image, which is primary
Scan obtained medical image.
Character data output module 1020 is used to export has the descriptive of mapping relations at least one medical image
Character data, the wherein descriptive character data include that at least one medical image itself is described by natural language
Data.
In accordance with an embodiment of the present disclosure, which may further include neural network input
Module and neural network export acquisition submodule.Neural network input submodule is used to input at least one medical image
To neural network.Neural network exports acquisition submodule, the output for obtaining the neural network, wherein the neural network
Output includes the descriptive character data.
It is understood that medical image acquisition module 1010 and character data output module 1020 may be incorporated in one
Realize that either any one module therein can be split into multiple modules or wherein any one module institute in module
Including multiple submodule can split or merge realization.Alternatively, one or more of these module and submodule mould
At least partly function of block can be combined at least partly function of other modules, and be realized in a module.According to this
The embodiment of invention, each submodule in medical image acquisition module 1010 and character data output module 1020 can be at least
It is implemented partly as hardware circuit, such as field programmable gate array (FPGA), programmable logic array (PLA), on piece system
System, the system on substrate, the system in encapsulation, application-specific integrated circuit (ASIC), or can be to be integrated or be encapsulated to circuit
The hardware such as any other rational method or firmware realize, or with software, three kinds of realization methods of hardware and firmware it is suitable
It is realized when combination.Alternatively, each submodule included by medical image acquisition module 1010 and character data output module 1020
It is in the block it is at least one can at least be implemented partly as computer program module, can when the program is run by computer
To execute the function of corresponding module.
Figure 11 diagrammatically illustrates the training method for handling medical image using robot according to the embodiment of the present disclosure
Flow chart.
As shown in figure 11, in accordance with an embodiment of the present disclosure, which includes
Operate S1101~operation S1103.
In operation S1101, at least one medical image is input to neural network to obtain the output of the neural network,
In, which is the medical image that single pass obtains, and the output of the neural network includes for describing this
The character data of at least one medical image.
In operation S1102, when the consistency of output and the model answer of the neural network does not meet preset condition, weight
Input operation is executed again, has been trained when the output of the neural network meets preset condition with the model answer consistency
At the wherein model answer includes and the predetermined descriptive character at least one medical image with mapping relations
Data, the descriptive character data include the number that described at least one medical image itself is described by natural language
According to.
In operation S1103, the neural network that training is completed is exported.
In accordance with an embodiment of the present disclosure, training neural network carries out the processing of medical image.In this way, refreshing net after the completion of training
Medical image can be translated as the descriptive character data by natural language description, such as text information etc. by network.To
Another way is provided for the reading of medical image.
Figure 12 diagrammatically illustrates the block diagram of the processing system 1200 according to the medical image of the embodiment of the present disclosure.
As shown in figure 12, include processor 1210 and meter according to the processing system 1200 of the medical image of the embodiment of the present disclosure
Calculation machine readable storage medium storing program for executing 1220.In some embodiments, the processing system 1200 of the medical image can execute above with reference to
The processing method of Figure 1B, Fig. 2 B, medical image described in Fig. 3 B and Fig. 4, correspondence is converted to realize by medical image
Descriptive character data.In further embodiments, the processing system 1200 of the medical image can be used for executing above
With reference to the method that figure 11 describes, medical image is handled with image training robot.
Specifically, processor 1210 for example may include general purpose microprocessor, instruction set processor and/or related chip group
And/or special microprocessor (for example, application-specific integrated circuit (ASIC)), etc..Processor 1210 can also include for caching
The onboard storage device of purposes.Processor 1210 can be performed for reference to described in figure 1B, Fig. 2 B, Fig. 3 B and Fig. 4
Single treatment unit either multiple processing units of the different actions of the process flow of medical image.
Computer readable storage medium 1220, such as can be able to include, store, transmit, propagate or transmit instruction
Arbitrary medium.For example, readable storage medium storing program for executing can include but is not limited to electricity, magnetic, optical, electromagnetic, infrared or semiconductor system, dress
It sets, device or propagation medium.The specific example of readable storage medium storing program for executing includes：Magnetic memory apparatus, such as tape or hard disk (HDD)；Light
Storage device, such as CD (CD-ROM)；Memory, such as random access memory (RAM) or flash memory；And/or wire/wireless communication
Link.
Computer readable storage medium 1220 may include computer program 1221, which may include
Code/computer executable instructions makes processor 1210 execute for example above in conjunction with figure when being executed by processor 1210
1B, Fig. 2 B, the process flow of medical image described in Fig. 3 B and Fig. 4 and its any deformation, such as above
It is described in conjunction with Figure 11 to handle the training method flow of medical image and its any deformation using robot.
Computer program 1221 can be configured with such as computer program code including computer program module.Example
Such as, in the exemplary embodiment, the code in computer program 1221 may include one or more program modules, such as including
1221A, module 1221B ....It should be noted that the dividing mode and number of module are not fixed, those skilled in the art
It can be combined using suitable program module or program module according to actual conditions, when these program modules are combined by processor
When 1210 execution so that processor 1210 can execute for example above in conjunction with described in Figure 1B, Fig. 2 B, Fig. 3 B and Fig. 4
The process flow of medical image and its any deformation, such as above in conjunction with method flow described in Figure 11 and its appoint
What is deformed.
According to an embodiment of the invention, medical image acquisition module 1010 and character data output module 1020 and wherein
Including each submodule at least one computer program module that can be implemented as describing with reference to figure 12, by processor
When 1210 execution, corresponding operating described above may be implemented.
It will be understood by those skilled in the art that the feature described in each embodiment and/or claim of the disclosure can
To carry out multiple combinations or/or combination, even if such combination or combination are not expressly recited in the disclosure.Particularly, exist
In the case of not departing from disclosure spirit or teaching, the feature described in each embodiment and/or claim of the disclosure can
To carry out multiple combinations and/or combination.All these combinations and/or combination each fall within the scope of the present disclosure.
Although the disclosure, art technology has shown and described with reference to the certain exemplary embodiments of the disclosure
Personnel it should be understood that in the case of the spirit and scope of the present disclosure limited without departing substantially from the following claims and their equivalents,
A variety of changes in form and details can be carried out to the disclosure.Therefore, the scope of the present disclosure should not necessarily be limited by above-described embodiment,
But should be not only determined by appended claims, also it is defined by the equivalent of appended claims.
Claims (10)
1. a kind of system of medical diagnosis, including：
One or more processors；And
Memory is stored with computer-readable instruction；Wherein, the processor is made when described instruction is executed by the processor
It realizes：
At least one medical image is obtained, at least one medical image is the medical image that single pass obtains；And
Output has the diagnosis report data of mapping relations at least one medical image.
2. system according to claim 1, wherein export has examining for mapping relations at least one medical image
Disconnected data reporting includes：
At least one medical image is input to neural network；And
Obtain the output of the neural network, wherein the output of the neural network includes the diagnosis report data.
3. system according to claim 2, wherein the neural network includes at least one first nerves network and at least
One nervus opticus network；At least one medical image, which is input to neural network, includes：
At least one medical image is input to the first nerves network to extract at least one medical image
At least one characteristic；And
At least one characteristic is input to the nervus opticus network, with by the nervus opticus network obtain with
At least one medical image has the diagnosis report data of mapping relations；
Wherein, the first nerves network and nervus opticus network are same type or different type.
4. system according to claim 3, wherein：
The first nerves network includes convolutional neural networks；And
The nervus opticus network includes Recognition with Recurrent Neural Network.
5. system according to claim 3, wherein when it is described at least a characteristic includes multiple characteristics when, will
At least one characteristic is input to the nervus opticus network, with by the nervus opticus network obtain with it is described extremely
A few medical image has the diagnosis report data of mapping relations, including：
It is a compressive features data by the multiple feature data compression；
The compressive features data are input to the nervus opticus network, with by the nervus opticus network obtain with it is described
At least one medical image has the diagnosis report data of mapping relations.
6. system according to claim 5, wherein by the multiple feature data compression be a compressive features data,
Including：
According to each characteristic and the current content status number of the nervus opticus network in the multiple characteristic
According to correlation determine that the compression weight of each characteristic in the multiple characteristic, wherein correlation get over high compression
Weight is bigger；
The multiple characteristic is weighted averagely, to obtain the compressive features data according to the compression weight.
7. a kind of device of medical diagnosis, including：
Medical image acquisition module, for obtaining at least one medical image, at least one medical image is single pass
Obtained medical image；And
Diagnosis report data outputting module, for exporting the diagnosis report that there are mapping relations at least one medical image
Data.
8. device according to claim 7, wherein the diagnosis report data outputting module includes：
Neural network input submodule, at least one medical image to be input to neural network；And
Neural network exports acquisition submodule, the output for obtaining the neural network, wherein the output of the neural network
Including the diagnosis report data.
9. device according to claim 8, wherein the neural network includes at least one first nerves network and at least
One nervus opticus network, the neural network input submodule include：
First nerves network inputs unit, at least one medical image to be input to the first nerves network to carry
Take at least one characteristic of at least one medical image；And
Nervus opticus network inputs unit, at least one characteristic to be input to the nervus opticus network, with
The diagnosis report data that there are mapping relations at least one medical image are obtained by the nervus opticus network；
Wherein：
The first nerves network and nervus opticus network are same type or different type.
10. a kind of training method carrying out medical diagnosis using robot, including：
At least one medical image is input to neural network to obtain the output of the neural network, wherein described at least one
A medical image is the medical image that single pass obtains, and the output of the neural network includes described at least one for describing
The character data of medical image；
When the consistency of output and the model answer of the neural network does not meet preset condition, the input is repeated
Operation, training completion, wherein institute when the output of the neural network meets preset condition with the model answer consistency
It includes the diagnosis report data for having mapping relations at least one medical image to state model answer；
The neural network that output training is completed.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2017114165041 | 2017-12-22 | ||
| CN201711416504 | 2017-12-22 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108399619A true CN108399619A (en) | 2018-08-14 |
| CN108399619B CN108399619B (en) | 2021-12-24 |
Family
ID=63091402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810171115.5A Active CN108399619B (en) | 2017-12-22 | 2018-03-01 | System and device for medical diagnosis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108399619B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021012225A1 (en) * | 2019-07-24 | 2021-01-28 | Beijing Didi Infinity Technology And Development Co., Ltd. | Artificial intelligence system for medical diagnosis based on machine learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150230773A1 (en) * | 2014-02-19 | 2015-08-20 | Samsung Electronics Co., Ltd. | Apparatus and method for lesion detection |
| CN106446782A (en) * | 2016-08-29 | 2017-02-22 | 北京小米移动软件有限公司 | Image identification method and device |
| CN106709254A (en) * | 2016-12-29 | 2017-05-24 | 天津中科智能识别产业技术研究院有限公司 | Medical diagnostic robot system |
| CN107045720A (en) * | 2017-05-04 | 2017-08-15 | 深圳硅基智能科技有限公司 | Artificial neural network and system for recognizing eye fundus image lesion |
| CN107123027A (en) * | 2017-04-28 | 2017-09-01 | 广东工业大学 | A kind of cosmetics based on deep learning recommend method and system |
| CN107239733A (en) * | 2017-04-19 | 2017-10-10 | 上海嵩恒网络科技有限公司 | Continuous hand-written character recognizing method and system |
-
2018
- 2018-03-01 CN CN201810171115.5A patent/CN108399619B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150230773A1 (en) * | 2014-02-19 | 2015-08-20 | Samsung Electronics Co., Ltd. | Apparatus and method for lesion detection |
| CN106446782A (en) * | 2016-08-29 | 2017-02-22 | 北京小米移动软件有限公司 | Image identification method and device |
| CN106709254A (en) * | 2016-12-29 | 2017-05-24 | 天津中科智能识别产业技术研究院有限公司 | Medical diagnostic robot system |
| CN107239733A (en) * | 2017-04-19 | 2017-10-10 | 上海嵩恒网络科技有限公司 | Continuous hand-written character recognizing method and system |
| CN107123027A (en) * | 2017-04-28 | 2017-09-01 | 广东工业大学 | A kind of cosmetics based on deep learning recommend method and system |
| CN107045720A (en) * | 2017-05-04 | 2017-08-15 | 深圳硅基智能科技有限公司 | Artificial neural network and system for recognizing eye fundus image lesion |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021012225A1 (en) * | 2019-07-24 | 2021-01-28 | Beijing Didi Infinity Technology And Development Co., Ltd. | Artificial intelligence system for medical diagnosis based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108399619B (en) | 2021-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hu et al. | Learning to reason: End-to-end module networks for visual question answering | |
| WO2020019738A1 (en) | Plaque processing method and device capable of performing magnetic resonance vessel wall imaging, and computing device | |
| CN109376843A (en) | EEG signals rapid classification method, implementation method and device based on FPGA | |
| CN107578099A (en) | Computing device and method | |
| Wu et al. | Dynamic filtering with large sampling field for convnets | |
| CN110880196B (en) | Tumor photoacoustic image rapid reconstruction method and device based on deep learning | |
| Qiang et al. | Modeling task-based fMRI data via deep belief network with neural architecture search | |
| US20190095793A1 (en) | Sensor quality upgrade framework | |
| CN115661336A (en) | Three-dimensional reconstruction method and related device | |
| Hosseininoorbin et al. | Scaling spectrogram data representation for deep learning on edge tpu | |
| CN114925320B (en) | Data processing method and related device | |
| CN108399619A (en) | The system and device of medical diagnosis | |
| CN110083842A (en) | Translation quality detection method, device, machine translation system and storage medium | |
| Hernandez-Cruz et al. | Neural style transfer as data augmentation for improving covid-19 diagnosis classification | |
| CN112101438B (en) | Left-right eye classification method, device, server and storage medium | |
| CN112348811A (en) | Registration method and device for deep learning image, electronic equipment and storage medium | |
| CN117218300A (en) | Three-dimensional model construction method, three-dimensional model construction training method and device | |
| Valencia et al. | A generalized hardware architecture for real-time spiking neural networks | |
| CN111652349A (en) | Neural network processing method and related equipment | |
| CN116229262A (en) | Modeling method and device for building information model | |
| CN116095183A (en) | Data compression method and related equipment | |
| CN110096145A (en) | Psychological condition display methods and device based on mixed reality and neural network | |
| CN109345545A (en) | A kind of method, apparatus and computer readable storage medium of segmented image generation | |
| Sinha et al. | Leapfrogging medical ai in low-resource contexts using edge tensor processing unit | |
| CN114897130A (en) | Deep learning model conversion and inference method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |