CN108877925A - A kind of determination method and device of pneumonia cause of disease - Google Patents
A kind of determination method and device of pneumonia cause of disease Download PDFInfo
- Publication number
- CN108877925A CN108877925A CN201810420019.XA CN201810420019A CN108877925A CN 108877925 A CN108877925 A CN 108877925A CN 201810420019 A CN201810420019 A CN 201810420019A CN 108877925 A CN108877925 A CN 108877925A
- Authority
- CN
- China
- Prior art keywords
- patient diagnosed
- pneumonia
- feature vector
- patient
- rabat image
- 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
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Radiology & Medical Imaging (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pathology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of determination method and devices of pneumonia cause of disease.The method includes：It is obtaining after the rabat image and analysis data of patient diagnosed, it can be obtained according to rabat image and analysis data to the corresponding feature vector of patient diagnosed, then default Classification Neural model can will be inputted to the corresponding feature vector of patient diagnosed, and according to the output result of the default Classification Neural model, it is determining after patient diagnosed is with pneumonia, can further obtain cause of disease type corresponding to the pneumonia suffered to patient diagnosed.In this way, on the one hand, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve the accuracy of pneumonia etiological diagnosis；On the other hand, the analysis data of patient is introduced, so as to determine the cause of disease type of pneumonia in terms of the rabat image of patient and analysis data two, effectively increases the confidence level of default Classification Neural model.
Description
This application claims submission on March 30th, 2018 Intellectual Property Right Bureau of the RPC, application No. is
201810289487.8, the Chinese patent application of entitled " a kind of determination method and device of pneumonia cause of disease " it is preferential
Power, entire contents are hereby incorporated by reference in the application.
Technical field
The present invention relates to field of medical technology more particularly to a kind of determination method and devices of pneumonia cause of disease.
Background technique
With getting worse for environmental pollution, more and more diseases show the trend of high outburst rate.Modern medical service skill
The development of art is very mature, and doctor can be diagnosed to be various diseases by medical knowledge and medical experience.That is, existing
There is the cause of disease of the determination disease in technology to rely on the diagnosis of doctor mostly, however since the medical level in each area is very different
It causes, and personal experience's level of doctor is also irregular, therefore, the method for traditional diagnosis disease is easy to be cured by area
Horizontal and doctor's personal experience's level influence is treated, the problem that Error Diagnostics are larger is caused.
By taking pneumonia as an example, doctor needs first to determine pneumonia cause of disease when treating pneumonia, to determine difference according to cause of disease
Remedy measures.And the cause of disease for causing patient to suffer from pneumonia has very much, such as virus, bacterium, Chlamydia, mycoplasma etc..Doctor
According to lung X-ray light image come the case where when artificially being diagnosed, inevitably will appear mistaken diagnosis.
Based on this, a kind of determination method of disease pathogen is needed at present, for improving the accuracy rate of diagnosis of pneumonia.
Summary of the invention
The embodiment of the present invention provides a kind of determination method and device of pneumonia cause of disease, to improve the accuracy rate of diagnosis of pneumonia.
The embodiment of the present invention provides a kind of determination method of pneumonia cause of disease, the method includes：
Obtain the rabat image and the analysis data to patient diagnosed to patient diagnosed；
According to the rabat image to patient diagnosed and the analysis data to patient diagnosed, obtain described wait make a definite diagnosis
The corresponding feature vector of patient；
Described default Classification Neural model will be inputted to the corresponding feature vector of patient diagnosed, and according to described default
The output of Classification Neural model obtains described to patient diagnosed's trouble as a result, described in determining after patient diagnosed is with pneumonia
Cause of disease type corresponding to some pneumonia；Wherein, the parameter of the default Classification Neural model be by it is multiple really
Examine the corresponding feature vector of patient, whether each patient diagnosed suffers from pneumonia and with cause of disease type corresponding to pneumonia into
Row training obtains.
So, on the one hand, for the mode of traditional diagnosis, the embodiment of the present invention uses preset classification
Neural network determines the cause of disease type of pneumonia, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve
The accuracy of pneumonia etiological diagnosis；On the other hand, the embodiment of the present invention introduces the analysis data of patient, so as to from patient
Rabat image and the aspect of analysis data two determine the cause of disease type of pneumonia, effectively increase default Classification Neural mould
The confidence level of type.
Optionally, according to the rabat image to patient diagnosed and the analysis data to patient diagnosed, institute is obtained
It states to the corresponding feature vector of patient diagnosed, including：
Feature extraction is carried out to the rabat image to patient diagnosed using default feature extraction neural network model, is obtained
To the corresponding first eigenvector of rabat image to patient diagnosed；The ginseng of the default feature extraction neural network model
Number is trained by the rabat image to the multiple patient diagnosed；And to described to patient diagnosed's
Analysis data is screened, and the corresponding second feature vector of analysis data to patient diagnosed is obtained；
The first eigenvector and the second feature vector are calculated by full articulamentum, obtained described to true
Examine the corresponding feature vector of patient.
It so, it is possible to effectively improve the accuracy of feature vector, and then improve the accuracy of pneumonia etiological diagnosis；And by
In the analysis data for introducing patient, therefore, pneumonia can be determined in terms of the rabat image of patient and analysis data two
Cause of disease type, to improve the confidence level of default Classification Neural model.
Optionally, the default feature extraction neural network includes N number of convolution module；N is less than or equal to first threshold.
Balanced rabat image and analysis data be so, it is possible to the influence degree of default disaggregated model, avoided due to rabat
Image causes influence of the feature extracted to subsequent default disaggregated model excessive by excessive convolution module, and then causes pre-
If the problem of the prediction result inaccuracy of the influence of disaggregated model.
Optionally, rabat image of the acquisition to patient diagnosed, including：
Obtain the initial rabat image to patient diagnosed；
Deformation process is carried out to the initial rabat image, obtains the rabat image to patient diagnosed；The deformation
Handle at least one including cutting, scaling, rotate, in brightness adjustment.
The picture size that so, it is possible to guarantee to input default feature extraction neural network model is consistent, image clearly, thus
Improve the accuracy of first eigenvector.
Optionally, the analysis data to patient diagnosed is by being chemically examined to patient diagnosed's progress blood routine described
The data arrived.
In this way, can preferably reflect function of human body due to carrying out the data that blood routine is chemically examined, it is normal using blood
Rule analysis data can be improved the disconnected accuracy of pneumonia disease original sentence.
The embodiment of the present invention provides a kind of determination device of pneumonia cause of disease, and described device includes：
Acquiring unit, for obtaining rabat image and the analysis data to patient diagnosed to patient diagnosed；
Processing unit, for according to the rabat image and the analysis data to patient diagnosed to patient diagnosed,
It obtains described to the corresponding feature vector of patient diagnosed；
The processing unit is also used to input default Classification Neural to the corresponding feature vector of patient diagnosed for described
Model, and according to the output of the default Classification Neural model as a result, being obtained described in determining after patient diagnosed is with pneumonia
To cause of disease type corresponding to the pneumonia suffered to patient diagnosed；Wherein, the ginseng of the default Classification Neural model
Number is by whether suffering from pneumonia and with pneumonia to the corresponding feature vector of multiple patients diagnosed, each patient diagnosed
What corresponding cause of disease type was trained.
Optionally, the processing unit is specifically used for：
Feature extraction is carried out to the rabat image to patient diagnosed using default feature extraction neural network model, is obtained
To the corresponding first eigenvector of rabat image to patient diagnosed；The ginseng of the default feature extraction neural network model
Number is trained by the rabat image to the multiple patient diagnosed；And to described to patient diagnosed's
Analysis data is screened, and the corresponding second feature vector of analysis data to patient diagnosed is obtained；And by described
One feature vector and the second feature vector are calculated by full articulamentum, are obtained described to the corresponding feature of patient diagnosed
Vector.
Optionally, the default feature extraction neural network includes N number of convolution module；N is less than or equal to first threshold.
Optionally, the acquiring unit is specifically used for：
Obtain the initial rabat image to patient diagnosed；And deformation process is carried out to the initial rabat image,
Obtain the rabat image to patient diagnosed；The deformation process include cut, scaling, rotation, in brightness adjustment at least
One.
Optionally, the analysis data to patient diagnosed is by being chemically examined to patient diagnosed's progress blood routine described
The data arrived.
In the embodiment of the present invention, obtaining after the rabat image and analysis data of patient diagnosed, it can be according to chest film picture
Picture and analysis data are obtained to the corresponding feature vector of patient diagnosed, then can will be inputted to the corresponding feature vector of patient diagnosed
Default Classification Neural model, and according to the output of the default Classification Neural model as a result, suffering from determining wait make a definite diagnosis
After person suffers from pneumonia, cause of disease type corresponding to the pneumonia suffered to patient diagnosed can be further obtained.In this way,
On the one hand, for the mode of traditional diagnosis, the embodiment of the present invention is using preset Classification Neural come really
The cause of disease type for determining pneumonia, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve pneumonia etiological diagnosis
Accuracy；On the other hand, the embodiment of the present invention introduces the analysis data of patient, so as to from the rabat image of patient and
The cause of disease type that pneumonia is determined in terms of analysis data two, effectively increases the confidence level of default Classification Neural model.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly introduced, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill in field, without any creative labor, it can also be obtained according to these attached drawings
His attached drawing.
Fig. 1 is flow diagram corresponding to a kind of determination method of pneumonia cause of disease provided in an embodiment of the present invention；
Fig. 2 is a kind of structural schematic diagram of default feature extraction neural network model provided in an embodiment of the present invention；
Fig. 3 is a kind of structural schematic diagram of default Classification Neural model provided in an embodiment of the present invention；
Fig. 4 is the structural schematic diagram of neural network model involved in the embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of the determination device of pneumonia cause of disease provided in an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention make into
It is described in detail to one step, it is clear that the described embodiments are only some of the embodiments of the present invention, rather than whole implementation
Example.Based on the embodiments of the present invention, obtained by those of ordinary skill in the art without making creative efforts
All other embodiment, shall fall within the protection scope of the present invention.
Fig. 1 illustrates process corresponding to a kind of determination method of pneumonia cause of disease provided in an embodiment of the present invention and shows
It is intended to, as shown in Figure 1, including the following steps：
Step 101, the rabat image and the analysis data to patient diagnosed to patient diagnosed are obtained
Step 102, according to the rabat image to patient diagnosed and the analysis data to patient diagnosed, institute is obtained
It states to the corresponding feature vector of patient diagnosed；
Step 103, it described will be inputted to the corresponding feature vector of patient diagnosed and preset Classification Neural model, and according to
The output of the default Classification Neural model obtains described to true as a result, described in determining after patient diagnosed is with pneumonia
Examine cause of disease type corresponding to the pneumonia that patient suffers from.
Using above-mentioned described method, on the one hand, for the mode of traditional diagnosis, the present invention is implemented
Example determines the cause of disease type of pneumonia, diagnosis caused by can reduce because of doctor's level difference using preset Classification Neural
Error rate, to improve the accuracy of pneumonia etiological diagnosis；On the other hand, the embodiment of the present invention introduces the chemical examination number of patient
According to effectively increasing so as to determine the cause of disease type of pneumonia in terms of the rabat image of patient and analysis data two
The confidence level of default Classification Neural model.
Specifically, in step 101, the rabat image to patient diagnosed can be the normotopia of chest X-ray to patient diagnosed
Picture.It, can be by the initial rabat image of patient, i.e., the normotopia of chest of the patient taken instrument in the embodiment of the present invention
Rabat image of the X-ray image directly as patient；Alternatively, can also be after the initial rabat image for obtaining patient, to described
Initial rabat image carries out deformation process, i.e., carries out deformation process to the normotopia of chest X-ray image of patient, and by deformation process
Rabat image of the image afterwards as patient.
Further, there are many modes that deformation process is carried out to normotopia of chest X-ray image, for example, can be to chest
Normotopia X-ray image takes the processing modes such as cutting, scaling, rotation, brightness adjustment to carry out deformation process.Specific implementation process
In, those skilled in the art can carry out deformation process to normotopia of chest X-ray image only with a kind of above-mentioned processing mode, or
Person can also carry out deformation process to normotopia of chest X-ray image using a variety of processing modes, specifically without limitation.As an example
The background color of image a can be first transformed to black by son after obtaining normotopia of chest X-ray image (indicating with image a)
Color obtains image b；Then the pixel value linear transformation of image b is obtained into image c to 0~255 this section；It then, can be with
Image c is zoomed in and out, for example scaling size is the 80% of original image, obtains image d；In turn, image d can be cut, than
Image d is such as cut to 3:4 size obtains image e, and finally obtained image e can be used as the rabat to patient diagnosed
Image.The picture size that so, it is possible to guarantee to input default feature extraction neural network model is consistent, image clearly, to mention
The accuracy of high first eigenvector.
It should be noted that above for example be only a kind of example, the application is to the successive suitable of each processing mode
Sequence is not specifically limited, for example, can be first by the pixel value linear transformation of image a to 0~255 this section, then by background face
Discoloration is changed to black, specifically without limitation.
Analysis data to patient diagnosed can be a variety of different types of data, for example, the chemical examination number to patient diagnosed
According to can be by carrying out the data (i.e. blood routine data) chemically examined of blood routine to patient diagnosed to described；Alternatively, wait make a definite diagnosis
The analysis data of patient can also be by carrying out the data that phlegm is chemically examined to patient diagnosed to described, specifically without limitation.
In step 102, to the corresponding feature vector of patient diagnosed method of determination there are many, a kind of possible implementation
For that can be obtained using default feature extraction neural network model to patient diagnosed's for the rabat image to patient diagnosed
The corresponding first eigenvector of rabat image can be according to the chemical examination to patient diagnosed for the analysis data to patient diagnosed
Data obtain the corresponding second feature vector of analysis data to patient diagnosed, and then can be special by first eigenvector and second
It levies vector to carry out after connecting entirely by full articulamentum, obtain to the corresponding feature vector of patient diagnosed.
In the embodiment of the present invention, the parameter of default feature extraction neural network model can be by it is the multiple really
Examine what the rabat image of patient was trained.Wherein, presetting feature extraction neural network model can be shallow-layer nerve net
Network model, i.e., the default feature extraction neural network may include N number of convolution module, and, N is less than or equal to first threshold.This
Field technical staff can rule of thumb set the specific value of first threshold with actual conditions, herein without limitation.
In order to which according to default feature extraction neural network model referred to above is clearly described, Fig. 2 is exemplary
Show a kind of structural schematic diagram of default feature extraction neural network model provided in an embodiment of the present invention.The default feature mentions
Take neural network model may include three convolution modules and be located at three convolution modules in the last one convolution mould
The full articulamentum of block connection.As shown in Fig. 2, three convolution modules are respectively the first convolution module 201, the second convolution module 202
With third convolution module 203, full articulamentum 204 can be connect with third convolution module 203；Wherein, each convolution module again may be used
To include ReLU that convolutional layer, the normalization connecting with convolutional layer (Batch Normalization, BN) layer, connect with BN layers
Activation primitive layer and with ReLU layers of pooling layers of the max connecting, the first convolution module 201 as shown in Figure 2 include first
Convolutional layer 2011, the first BN layer 2012, the first ReLU activation primitive layer 2013 and the first max pooling layer 2014, volume Two
Volume module 202 includes the second convolutional layer 2021, the 2nd BN layer 2022, the 2nd ReLU activation primitive layer 2023 and the 2nd max
Pooling layer 2024, third convolution module 203 include third convolutional layer 2031, the 3rd BN layer 2032, the 3rd ReLU activation primitive
Layer 2033 and the 3rd max pooling layer 2034.
It should be noted that the convolution kernel size of each convolutional layer shown in Figure 2, pooling layers of max of convolution kernel
The feature channel numerical value that size, each convolution module are extracted can rule of thumb be set with actual conditions for those skilled in the art
And adjust and obtain, specifically without limitation.
It further, can be by the chest of multiple patients diagnosed before the default feature extraction neural network model of training
Picture carries out data enhancing, to expand the training sample amount of default feature extraction neural network model, and then improves default
The accuracy of feature extraction neural network model.In the embodiment of the present invention, there are many modes of data enhancing, for example, can be right
The rabat image of patient diagnosed carries out data enhancing by the way of Random Level mirror image；Alternatively, can also suffer to having made a definite diagnosis
The rabat image of person carries out data enhancing in such a way that random upper and lower, left and right translate；Alternatively, can also be to patient diagnosed
Rabat image data enhancing is carried out by the way of Random-Rotation；Alternatively, can also the rabat image to patient diagnosed adopt
Data enhancing is carried out with the mode scaled at random, specifically without limitation.
In the embodiment of the present invention, the method for determination of the corresponding second feature vector of analysis data to patient diagnosed has more
Kind, in an example, can treat patient diagnosed analysis data screen after, obtain the chemical examination number to patient diagnosed
According to corresponding second feature vector.It is blood routine data instance with analysis data, as shown in table 1, for the blood routine number before screening
According to a kind of example.
Table 1：A kind of example of blood routine data before screening
| Blood routine project | Numerical value |
| White blood cell count(WBC) | 6*10^9/L |
| Neutrophil leucocyte percentage | 65% |
| Cent lymphocytes | 33% |
| Monocyte percentage | 5% |
| Lymphocyte count | 2*10^9/L |
| Monocyte count | 0.2*10^9/L |
| Red blood cell count(RBC) | 4*10^9/L |
Blood routine based on above content, after being screened to blood routine data shown in table 1, after available screening
Data, it is specific as shown in table 2.According to content shown in table 2 it is found that second feature vector can be [65,33,5].
Table 2：A kind of example of blood routine data after screening
| Blood routine project | Numerical value |
| Neutrophil leucocyte percentage | 65% |
| Cent lymphocytes | 33% |
| Monocyte percentage | 5% |
It should be noted that the content shown in table 2 is only a kind of possible example, those skilled in the art can be according to warp
It tests the blood routine data for obtaining inspection with actual conditions to screen, that is to say, that those skilled in the art can be according to warp
It tests and additions and deletions is carried out to content shown in table 2 with actual conditions, specifically without limitation.
In other examples, it can also will directly be shown in table 1 directly using blood routine data as second feature vector
Content as second feature vector, that is to say, that according to content shown in table 1, second feature vector may be [6,
65,33,5,2,0.2,4].
It further, can be using full articulamentum to described after determining first eigenvector and second feature vector
First eigenvector and the second feature vector are calculated, to obtain to the corresponding feature vector of patient diagnosed.It lifts a
Example, if first eigenvector is the feature vector of M dimension, second feature vector is the feature vector of N-dimensional, then uses full articulamentum
After calculating the first eigenvector and the second feature vector, obtain to the corresponding feature vector of patient diagnosed
For the feature vector of M+N dimension.
Method of determination as described above to the corresponding feature vector of patient diagnosed is only a kind of possible implementation,
In other possible implementations, it can also determine otherwise to the corresponding feature vector of patient diagnosed, specifically not
It limits.
In step 103, the parameter of default Classification Neural model be can be by corresponding to multiple patients diagnosed
Whether feature vector, each patient diagnosed suffer from pneumonia and are trained to obtain with cause of disease type corresponding to pneumonia
's.
Fig. 3 illustrates a kind of structural representation of default Classification Neural model provided in an embodiment of the present invention
Figure, as shown in figure 3, the default Classification Neural model includes the first full articulamentum 301, the second complete 302 and of articulamentum
Softmax layer 303.The first full articulamentum 301, the second full articulamentum can be passed sequentially through to the corresponding feature vector of patient diagnosed
After 302 are calculated, then output category result after being classified by softmax layer 303, so that it is determined that whether suffering to patient diagnosed
There is pneumonia, and described after patient diagnosed is with pneumonia determining, obtains corresponding to the pneumonia suffered to patient diagnosed
Cause of disease type.
According to content as described above, Fig. 3 shows the structure of default Classification Neural model, but pre- for this
If the parameter of Classification Neural model, the embodiment of the present invention needs to show Fig. 3 using the pneumonia information of multiple patients diagnosed
Default Classification Neural model out can just obtain after being trained.In order to clearly describe default Classification Neural
The training process of model, below by for example bright.
It as shown in table 3, is a kind of example of the corresponding pneumonia information of patient diagnosed.The feature vector of patient diagnosed 1
For X1, pneumonia is suffered from, and the cause of disease type of pneumonia is virus；The feature vector of patient diagnosed 2 is X2, suffer from pneumonia, and pneumonia
Cause of disease type be bacterium；The feature vector of patient diagnosed 3 is X3, do not suffer from pneumonia；The feature vector of patient diagnosed 4 is
X4, pneumonia is suffered from, and the cause of disease type of pneumonia is mycoplasma；The feature vector of patient diagnosed 5 is X5, suffer from pneumonia, and pneumonia
Cause of disease type be Chlamydia.
Table 3：A kind of example of the corresponding pneumonia information of patient diagnosed
| Number | Feature vector | Whether pneumonia is suffered from | The corresponding cause of disease type of pneumonia |
| Patient diagnosed 1 | X1 | It is | Virus |
| Patient diagnosed 2 | X2 | It is | Bacterium |
| Patient diagnosed 3 | X3 | It is no | — |
| Patient diagnosed 4 | X4 | It is | Mycoplasma |
| Patient diagnosed 5 | X5 | It is | Chlamydia |
| …… | …… | …… | …… |
Further, whether the feature vector of multiple patients diagnosed and each patient shown in table 3 are suffered from into lung
Inflammation inputs in default Classification Neural model with cause of disease type corresponding to pneumonia, can determine default classification nerve net
The parameter of network model.For example, shown in the table 3 for patient diagnosed 1, the corresponding feature vector of patient diagnosed 1 is defeated
Enter in default Classification Neural, by propagated forward, the result vector of available one 5 dimensionWherein, y1It is right
The result answered is not suffer from pneumonia；y2Corresponding result is with pneumonia, and the cause of disease type of pneumonia is virus；y3Corresponding knot
Fruit is with pneumonia, and the cause of disease type of pneumonia is bacterium；y4Corresponding result is with pneumonia, and the cause of disease type of pneumonia is
Mycoplasma；y5Corresponding result is with pneumonia, and the cause of disease type of pneumonia is Chlamydia.In this way, if y in vector Y3Value most
Greatly, then illustrate that the determining patient diagnosed's 1 of default Classification Neural model suffers from pneumonia, and the cause of disease type of pneumonia is thin
Bacterium.And in fact, patient diagnosed 1 has suffered from pneumonia, and the cause of disease type of pneumonia is virus, Classification Neural mould default in this way
It there is error between the prediction result and actual result of type, i.e. loss (loss) functional value.In turn, backpropagation can be used
Algorithm, according to stochastic gradient descent (Stochastic gradient descent, SGD) algorithm along loss (loss) function
The direction of value decline adjusts the parameter of default Classification Neural model.
In this way, the cause of disease type of pneumonia is determined by presetting Classification Neural model, compared to traditional diagnosis
Mode for, the embodiment of the present invention determines the cause of disease type of pneumonia using preset Classification Neural, can reduce because
Error Diagnostics rate caused by doctor's level difference, to improve the accuracy of pneumonia etiological diagnosis.
In order to clearly describe the determination method of pneumonia cause of disease described above, below to involved by the embodiment of the present invention
And process carry out globality explanation.As shown in figure 4, for the structural representation of neural network model involved in the embodiment of the present invention
Figure.Go out structure according to Fig.4, it is found that can input the rabat image pre- obtaining after the rabat image of patient diagnosed
If Classification Neural model, to extract the corresponding first eigenvector of rabat image；Further, it is obtaining wait make a definite diagnosis trouble
After the analysis data of person, which can be screened, to obtain the corresponding second feature vector of analysis data；More
Further, first eigenvector and second feature vector can be calculated by full articulamentum, to obtain suffering from wait make a definite diagnosis
The feature vector of person, and the feature vector to patient diagnosed is inputted into default Classification Neural model, and then can treat really
It examines patient and whether suffers from pneumonia and analyze, and determining after patient diagnosed is with pneumonia, obtain suffering to patient diagnosed
Cause of disease type corresponding to pneumonia.
Based on same inventive concept, a kind of determination device of pneumonia cause of disease provided in an embodiment of the present invention, if Fig. 5 shows,
The device includes acquiring unit 501, processing unit 502；Wherein,
Acquiring unit 501, for obtaining rabat image and the analysis data to patient diagnosed to patient diagnosed；
Processing unit 502, for according to the rabat image and the chemical examination number to patient diagnosed to patient diagnosed
According to obtaining described to the corresponding feature vector of patient diagnosed；
The processing unit 502 is also used to described to the default classification nerve of the corresponding feature vector input of patient diagnosed
Network model, and according to the output of the default Classification Neural model as a result, suffering from pneumonia to patient diagnosed described in determining
Afterwards, cause of disease type corresponding to the pneumonia suffered to patient diagnosed is obtained；Wherein, the default Classification Neural model
Parameter be by the way that whether the corresponding feature vector of multiple patients diagnosed, each patient diagnosed are suffered from pneumonia and suffered from
Cause of disease type corresponding to pneumonia is trained.
Optionally, the processing unit 502 is specifically used for：
Feature extraction is carried out to the rabat image to patient diagnosed using default feature extraction neural network model, is obtained
To the corresponding first eigenvector of rabat image to patient diagnosed；The ginseng of the default feature extraction neural network model
Number is trained by the rabat image to the multiple patient diagnosed；And to described to patient diagnosed's
Analysis data is screened, and the corresponding second feature vector of analysis data to patient diagnosed is obtained；And by described
One feature vector and the second feature vector are calculated by full articulamentum, are obtained described to the corresponding feature of patient diagnosed
Vector.
Optionally, the default feature extraction neural network includes N number of convolution module；N is less than or equal to first threshold.
Optionally, the acquiring unit 501 is specifically used for：
Obtain the initial rabat image to patient diagnosed；And deformation process is carried out to the initial rabat image,
Obtain the rabat image to patient diagnosed；The deformation process include cut, scaling, rotation, in brightness adjustment at least
One.
Optionally, the analysis data to patient diagnosed is by being chemically examined to patient diagnosed's progress blood routine described
The data arrived.
In the embodiment of the present invention, obtaining after the rabat image and analysis data of patient diagnosed, it can be according to chest film picture
Picture and analysis data are obtained to the corresponding feature vector of patient diagnosed, then can will be inputted to the corresponding feature vector of patient diagnosed
Default Classification Neural model, and according to the output of the default Classification Neural model as a result, suffering from determining wait make a definite diagnosis
After person suffers from pneumonia, cause of disease type corresponding to the pneumonia suffered to patient diagnosed can be further obtained.In this way,
On the one hand, for the mode of traditional diagnosis, the embodiment of the present invention is using preset Classification Neural come really
The cause of disease type for determining pneumonia, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve pneumonia etiological diagnosis
Accuracy；On the other hand, the embodiment of the present invention introduces the analysis data of patient, so as to from the rabat image of patient and
The cause of disease type that pneumonia is determined in terms of analysis data two, effectively increases the confidence level of default Classification Neural model.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of determination method of pneumonia cause of disease, which is characterized in that the method includes：
Obtain the rabat image and the analysis data to patient diagnosed to patient diagnosed；
According to the rabat image to patient diagnosed and the analysis data to patient diagnosed, obtain described to patient diagnosed
Corresponding feature vector；
Described default Classification Neural model will be inputted to the corresponding feature vector of patient diagnosed, and according to the default classification
The output of neural network model obtains described suffering to patient diagnosed as a result, determine described after patient diagnosed is with pneumonia
Cause of disease type corresponding to pneumonia；Wherein, the parameter of the default Classification Neural model is by having made a definite diagnosis trouble to multiple
Whether the corresponding feature vector of person, each patient diagnosed suffer from pneumonia and instruct with cause of disease type corresponding to pneumonia
It gets.
2. the method according to claim 1, wherein according to the rabat image to patient diagnosed and it is described to
The analysis data of patient diagnosed, obtain it is described to the corresponding feature vector of patient diagnosed, including：
Feature extraction is carried out to the rabat image to patient diagnosed using default feature extraction neural network model, obtains institute
State the corresponding first eigenvector of rabat image to patient diagnosed；The parameter of the default feature extraction neural network model is
It is trained by the rabat image to the multiple patient diagnosed；And to the chemical examination to patient diagnosed
Data are screened, and the corresponding second feature vector of analysis data to patient diagnosed is obtained；
The first eigenvector and the second feature vector are calculated by full articulamentum, obtained described wait make a definite diagnosis trouble
The corresponding feature vector of person.
3. according to the method described in claim 2, it is characterized in that, the default feature extraction neural network includes N number of convolution
Module；N is less than or equal to first threshold.
4. the method according to claim 1, wherein the rabat image of the acquisition to patient diagnosed, including：
Obtain the initial rabat image to patient diagnosed；
Deformation process is carried out to the initial rabat image, obtains the rabat image to patient diagnosed；The deformation process
Including at least one in cutting, scaling, rotation, brightness adjustment.
5. method according to claim 1 to 4, which is characterized in that the analysis data to patient diagnosed
For by carrying out the data that blood routine is chemically examined to patient diagnosed to described.
6. a kind of determination device of pneumonia cause of disease, which is characterized in that described device includes：
Acquiring unit, for obtaining rabat image and the analysis data to patient diagnosed to patient diagnosed；
Processing unit, for obtaining according to the rabat image and the analysis data to patient diagnosed to patient diagnosed
It is described to the corresponding feature vector of patient diagnosed；
The processing unit is also used to input default Classification Neural mould to the corresponding feature vector of patient diagnosed for described
Type, and according to the output of the default Classification Neural model as a result, being obtained described in determining after patient diagnosed is with pneumonia
Cause of disease type corresponding to the pneumonia suffered to patient diagnosed；Wherein, the parameter of the default Classification Neural model
It is by whether suffering from pneumonia and with pneumonia institute to the corresponding feature vector of multiple patients diagnosed, each patient diagnosed
What corresponding cause of disease type was trained.
7. device according to claim 6, which is characterized in that the processing unit is specifically used for：
Feature extraction is carried out to the rabat image to patient diagnosed using default feature extraction neural network model, obtains institute
State the corresponding first eigenvector of rabat image to patient diagnosed；The parameter of the default feature extraction neural network model is
It is trained by the rabat image to the multiple patient diagnosed；And to the chemical examination to patient diagnosed
Data are screened, and the corresponding second feature vector of analysis data to patient diagnosed is obtained；And it is special by described first
Sign vector sum described in second feature vector calculated by full articulamentum, obtain described in the corresponding feature of patient diagnosed to
Amount.
8. device according to claim 7, which is characterized in that the default feature extraction neural network includes N number of convolution
Module；N is less than or equal to first threshold.
9. device according to claim 6, which is characterized in that the acquiring unit is specifically used for：
Obtain the initial rabat image to patient diagnosed；And deformation process is carried out to the initial rabat image, it obtains
The rabat image to patient diagnosed；The deformation process include cut, scaling, rotation, in brightness adjustment at least one of.
10. the device according to any one of claim 6 to 7, which is characterized in that the analysis data to patient diagnosed
For by carrying out the data that blood routine is chemically examined to patient diagnosed to described.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2018102894878 | 2018-03-30 | ||
| CN201810289487 | 2018-03-30 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108877925A true CN108877925A (en) | 2018-11-23 |
Family
ID=64327586
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810420019.XA Pending CN108877925A (en) | 2018-03-30 | 2018-05-04 | A kind of determination method and device of pneumonia cause of disease |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108877925A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852377A (en) * | 2019-11-08 | 2020-02-28 | 湘潭大学 | Method for identifying pneumonia according to lung X-ray film |
| CN111599465A (en) * | 2020-05-13 | 2020-08-28 | 上海森亿医疗科技有限公司 | Method, device, terminal and medium for predicting etiology type of children community acquired pneumonia |
| CN111932492A (en) * | 2020-06-24 | 2020-11-13 | 数坤（北京）网络科技有限公司 | Medical image processing method and device and computer readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130197384A1 (en) * | 2012-02-01 | 2013-08-01 | National Tsing Hua University | Medical Ventilator Capable of Early Detecting and Recognizing Types of Pneumonia, Gas Recognition Chip, and Method for Recognizing Gas Thereof |
| CN106355035A (en) * | 2016-09-28 | 2017-01-25 | 湖南老码信息科技有限责任公司 | Pneumonia prediction method and prediction system based on incremental neural network model |
| CN106951710A (en) * | 2017-03-22 | 2017-07-14 | 华东师范大学 | CAP data systems and method based on privilege information Learning support vector machine |
| US20170337686A1 (en) * | 2016-05-19 | 2017-11-23 | Sichuan University | Kind of x-ray chest image rib suppression method based on poisson model |
-
2018
- 2018-05-04 CN CN201810420019.XA patent/CN108877925A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130197384A1 (en) * | 2012-02-01 | 2013-08-01 | National Tsing Hua University | Medical Ventilator Capable of Early Detecting and Recognizing Types of Pneumonia, Gas Recognition Chip, and Method for Recognizing Gas Thereof |
| US20170337686A1 (en) * | 2016-05-19 | 2017-11-23 | Sichuan University | Kind of x-ray chest image rib suppression method based on poisson model |
| CN106355035A (en) * | 2016-09-28 | 2017-01-25 | 湖南老码信息科技有限责任公司 | Pneumonia prediction method and prediction system based on incremental neural network model |
| CN106951710A (en) * | 2017-03-22 | 2017-07-14 | 华东师范大学 | CAP data systems and method based on privilege information Learning support vector machine |
Non-Patent Citations (2)
| Title |
|---|
| 相文波: "基于卷积神经网络的肺炎类型影像判别", 《中国优秀硕士学位论文全文数据库》 * |
| 耿欢等: ""基于CT 影像的肺组织分割及其功能定量分析"", 《小型微型计算机系统》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852377A (en) * | 2019-11-08 | 2020-02-28 | 湘潭大学 | Method for identifying pneumonia according to lung X-ray film |
| CN111599465A (en) * | 2020-05-13 | 2020-08-28 | 上海森亿医疗科技有限公司 | Method, device, terminal and medium for predicting etiology type of children community acquired pneumonia |
| CN111932492A (en) * | 2020-06-24 | 2020-11-13 | 数坤（北京）网络科技有限公司 | Medical image processing method and device and computer readable storage medium |
| CN111932492B (en) * | 2020-06-24 | 2021-05-11 | 数坤（北京）网络科技有限公司 | Medical image processing method and device and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200340909A1 (en) | Method, apparatus, and computer program for supporting disease analysis, and method, apparatus, and program for training computer algorithm | |
| US20190065897A1 (en) | Medical image analysis method, medical image analysis system and storage medium | |
| CN108346145A (en) | The recognition methods of unconventional cell in a kind of pathological section | |
| CN108648179A (en) | A kind of method and device of analysis Lung neoplasm | |
| CN104573742B (en) | Classification method of medical image and system | |
| CN104484886B (en) | A kind of dividing method and device of MR images | |
| CN108629764A (en) | A kind of good pernicious method and device of determining Lung neoplasm | |
| CN109508644A (en) | Facial paralysis grade assessment system based on the analysis of deep video data | |
| CN108877925A (en) | A kind of determination method and device of pneumonia cause of disease | |
| US11756199B2 (en) | Image analysis in pathology | |
| JP2023544466A (en) | Training method and device for diagnostic model of lung adenocarcinoma and squamous cell carcinoma based on PET/CT | |
| CN110991254B (en) | Ultrasonic image video classification prediction method and system | |
| CN113782184A (en) | Cerebral apoplexy auxiliary evaluation system based on facial key point and feature pre-learning | |
| CN109447088A (en) | A kind of method and device of breast image identification | |
| CN111524594A (en) | Target population blood system malignant tumor screening system | |
| CN111062953A (en) | Method for identifying parathyroid hyperplasia in ultrasonic image | |
| CN116579982A (en) | Pneumonia CT image segmentation method, device and equipment | |
| CN111260667A (en) | Neurofibroma segmentation method combined with space guidance | |
| CN114494263B (en) | Medical image lesion detection method, system and equipment integrating clinical information | |
| CN114742802A (en) | Pancreas CT image segmentation method based on 3D transform mixed convolution neural network | |
| CN113269737A (en) | Method and system for calculating diameter of artery and vein of retina | |
| Yang et al. | A multi-stage progressive learning strategy for COVID-19 diagnosis using chest computed tomography with imbalanced data | |
| WO2024074921A1 (en) | Distinguishing a disease state from a non-disease state in an image | |
| CN112036298A (en) | Cell detection method based on double-segment block convolutional neural network | |
| CN112365962B (en) | Medical resource allocation method and equipment based on lung CT image and clinical information |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20190620 Address after: 310000 Room 177, Building No. 16, Zhuantang Science and Technology Economic Zone, Xihu District, Hangzhou City, Zhejiang Province Applicant after: According to the Hangzhou Medical Technology Co., Ltd. Address before: 310000 Room 177, Building No. 16, Zhuantang Science and Technology Economic Zone, Xihu District, Hangzhou City, Zhejiang Province Applicant before: According to the Hangzhou Medical Technology Co., Ltd. Applicant before: Hangzhou Yi chart Network Technology Co., Ltd. Applicant before: Guangzhou Yi chart Medical Technology Co., Ltd. |
|
| TA01 | Transfer of patent application right | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181123 |
|
| RJ01 | Rejection of invention patent application after publication |