CN109523546A - A kind of method and device of Lung neoplasm analysis - Google Patents
A kind of method and device of Lung neoplasm analysis Download PDFInfo
- Publication number
- CN109523546A CN109523546A CN201811570841.0A CN201811570841A CN109523546A CN 109523546 A CN109523546 A CN 109523546A CN 201811570841 A CN201811570841 A CN 201811570841A CN 109523546 A CN109523546 A CN 109523546A
- Authority
- CN
- China
- Prior art keywords
- lung neoplasm
- sign
- lung
- patient
- default
- 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
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- 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
- G06T2207/30064—Lung nodule
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a kind of method and devices of Lung neoplasm analysis, this method includes the three-dimensional coordinate of Lung neoplasm in the Lung neoplasm image and Lung neoplasm image for obtain the lung of patient, the segmentation image of lung, the segmentation image of Lung neoplasm, the ROI comprising the Lung neoplasm is determined from Lung neoplasm image according to the three-dimensional coordinate of Lung neoplasm, by the ROI of Lung neoplasm, the three-dimensional coordinate of Lung neoplasm, the segmentation image of lung and the segmentation image of Lung neoplasm are input to default feature extraction neural network model and carry out feature extraction, obtain the feature vector of Lung neoplasm image, the feature vector of Lung neoplasm image is input to default Classification Neural model, obtain the Lung neoplasm analysis result of the patient of default Classification Neural model output.For the mode of traditional diagnosis, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve the accuracy of Lung neoplasm analysis.
Description
Technical field
The present embodiments relate to the method and devices that machine learning techniques field more particularly to a kind of Lung neoplasm are analyzed.
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 most diagnosis by doctor of the determination disease in technology, however since the medical level in each area is very inconsistent, and
Personal experience's level of doctor is also irregular, and therefore, the method for traditional diagnosis disease is easy by regional healthcare water
Flat and doctor's personal experience's level image, leads to the problem that Error Diagnostics are larger.
By taking Lung neoplasm as an example, doctor usually requires artificially to observe lung image, with Lung neoplasm that patient is suffered from into
The case where row analysis, this process inevitably will appear mistaken diagnosis.
Based on this, a kind of method for analyzing Lung neoplasm is needed at present, for improving the accuracy rate of analysis Lung neoplasm.
Summary of the invention
The embodiment of the present invention provides a kind of method and device of Lung neoplasm analysis, to improve the accurate of analysis Lung neoplasm
Rate.
A kind of method of Lung neoplasm analysis provided in an embodiment of the present invention, comprising:
Obtain three-dimensional coordinate, the lung of Lung neoplasm in the Lung neoplasm image and the Lung neoplasm image of the lung of patient
The segmentation image for dividing image, the Lung neoplasm in portion；
It is determined from the Lung neoplasm image according to the three-dimensional coordinate of the Lung neoplasm interested comprising the Lung neoplasm
Region (Region of Interest, abbreviation ROI)；
By the ROI of the Lung neoplasm, the three-dimensional coordinate of the Lung neoplasm, the lung segmentation image and the Lung neoplasm
Segmentation image be input to default feature extraction neural network model and carry out feature extraction, obtain the feature of the Lung neoplasm image
Vector；The default feature extraction neural network model is by marked Lung neoplasm image, marked Lung neoplasm
Three-dimensional coordinate, the segmentation image of marked lung, marked Lung neoplasm segmentation image be trained；
The feature vector of the Lung neoplasm image is input to default Classification Neural model, obtains the default classification
The Lung neoplasm of the patient of neural network model output analyzes result；Wherein, the default Classification Neural model is logical
It crosses and the feature vector of the Lung neoplasm image of multiple patients and patient is trained with the known results of Lung neoplasm.
In this way, the embodiment of the present invention using default feature extraction neural network model to the ROI of Lung neoplasm, Lung neoplasm three
The segmentation Extraction of Image feature vector for dividing image and Lung neoplasm of coordinate, lung is tieed up, and is come in conjunction with default Classification Neural
Lung neoplasm is analyzed, since model above is by being trained to mass data, so that passing through mould
The result that type obtains is relatively reasonable, and has certain scientific basis.It, can for the mode of traditional diagnosis
Error Diagnostics rate caused by reducing because of doctor's level difference, to improve the accuracy of Lung neoplasm analysis.
Optionally, the default Classification Neural model is default sign Classification Neural model；The default sign
As Classification Neural model be by the sign of the Lung neoplasm that the corresponding feature vector of multiple patients, each patient are suffered from into
Row training obtains；
The Lung neoplasm analysis result of the patient of the default Classification Neural model output is that the patient suffers from
Lung neoplasm sign.
In this way, for doctor judges the sign of Lung neoplasm by way of artificially diagnosing in compared with the prior art, this
Inventive embodiments judge Lung neoplasm using default feature extraction neural network model, and in conjunction with default sign Classification Neural
Sign, Error Diagnostics rate caused by can reduce because of doctor's level difference, to improve the accuracy of determining Lung neoplasm sign.
Optionally, the default Classification Neural model is the feature vector by the Lung neoplasm image to multiple patients
And patient is trained with the known sign result of Lung neoplasm, comprising:
The corresponding feature vector of the multiple patient is input to initial sign Classification Neural model, is obtained each
The prediction sign for the Lung neoplasm that patient suffers from；
The reality for the Lung neoplasm that the prediction sign of the Lung neoplasm suffered from according to each patient and each patient suffer from
Sign carries out reverse train, generates the default sign Classification Neural model.
In this way, can accurately adjust default sign classification nerve by the comparison between prediction sign and practical sign
The parameter of network model improves the accuracy of the default sign Classification Neural model of generation.
Optionally, described that the corresponding feature vector of the patient is inputted into default Classification Neural model, and obtain institute
State the analysis result for the Lung neoplasm that the patient that default Classification Neural model exports suffers from, comprising:
The corresponding feature vector of the patient is inputted into default Classification Neural model, obtains that default sign is corresponding sets
Reliability；
If the corresponding confidence level of the default sign is greater than preset threshold, suffer from using the default sign as the patient
The sign of some Lung neoplasms.
In this way, confidence level, is greater than the sign of preset threshold by the comparison of confidence level and preset threshold by each sign
As the sign for the Lung neoplasm that patient suffers from, so as to effectively avoid ignoring the same Lung neoplasm, there are the feelings of a variety of signs
Condition.
Optionally, the sign is pleural indentation sign.
Correspondingly, the embodiment of the invention also provides a kind of devices of Lung neoplasm analysis, comprising:
Acquiring unit, the three of Lung neoplasm in the Lung neoplasm image and the Lung neoplasm image for obtaining the lung of patient
Tie up the segmentation image of coordinate, the segmentation image of the lung, the Lung neoplasm；
Determination unit determines to include the lung for the three-dimensional coordinate according to the Lung neoplasm from the Lung neoplasm image
The ROI of tubercle；
Processing unit, for by the segmentation shadow of the three-dimensional coordinate of the ROI of the Lung neoplasm, the Lung neoplasm, the lung
The segmentation image of picture and the Lung neoplasm is input to default feature extraction neural network model and carries out feature extraction, obtains the lung
The feature vector of tubercle image；The default feature extraction neural network model be by marked Lung neoplasm image,
The three-dimensional coordinate of the Lung neoplasm of label, the segmentation image of marked lung, marked Lung neoplasm segmentation image instructed
It gets；And the feature vector of the Lung neoplasm image is input to default Classification Neural model, it obtains described pre-
If the Lung neoplasm of the patient of Classification Neural model output analyzes result；Wherein, the default Classification Neural mould
Type is to be trained by the feature vector of the Lung neoplasm image to multiple patients and patient with the known results of Lung neoplasm
It obtains.
Optionally, the default Classification Neural model is default sign Classification Neural model；The default sign
As Classification Neural model be by the sign of the Lung neoplasm that the corresponding feature vector of multiple patients, each patient are suffered from into
Row training obtains；
The Lung neoplasm analysis result of the patient of the default Classification Neural model output is that the patient suffers from
Lung neoplasm sign.
Optionally, the processing unit is specifically used for:
The corresponding feature vector of the multiple patient is input to initial sign Classification Neural model, is obtained each
The prediction sign for the Lung neoplasm that patient suffers from；
The reality for the Lung neoplasm that the prediction sign of the Lung neoplasm suffered from according to each patient and each patient suffer from
Sign carries out reverse train, generates the default sign Classification Neural model.
Optionally, the processing unit is specifically used for:
The corresponding feature vector of the patient is inputted into default Classification Neural model, obtains that default sign is corresponding sets
Reliability；
If the corresponding confidence level of the default sign is greater than preset threshold, suffer from using the default sign as the patient
The sign of some Lung neoplasms.
Optionally, the sign is pleural indentation sign.
Correspondingly, the embodiment of the invention also provides a kind of computer readable storage medium, the computer-readable storage
Media storage has computer executable instructions, and the computer executable instructions are for making the computer execute above-mentioned Lung neoplasm
The method of analysis.
Correspondingly, the embodiment of the invention also provides a kind of calculating equipment of image recognition, comprising:
Memory, for storing program instruction；
Processor executes above-mentioned lung knot according to the program of acquisition for calling the program instruction stored in the memory
The method for saving analysis.
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 creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of schematic diagram of system architecture provided in an embodiment of the present invention；
Fig. 2 is a kind of flow diagram of the method for Lung neoplasm analysis provided in an embodiment of the present invention；
Fig. 3 is a kind of schematic diagram of Lung neoplasm image provided in an embodiment of the present invention；
Fig. 4 a and Fig. 4 b are a kind of schematic diagram of Lung neoplasm image provided in an embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of default feature extraction neural network model provided in an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of default Classification Neural model provided in an embodiment of the present invention；
Fig. 7 is a kind of structural schematic diagram of the device of Lung neoplasm analysis provided in an embodiment of the present invention；
Fig. 8 is a kind of structural schematic diagram of the equipment of Lung neoplasm analysis 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 described embodiments are only a part 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.
The applicable system architecture of the method that Fig. 1 is analyzed by Lung neoplasm provided in an embodiment of the present invention.Refering to what is shown in Fig. 1,
The system architecture can be server 100, including processor 110, communication interface 120 and memory 130.
Wherein, the terminal device that communication interface 120 is applicable in for doctor communicates, and receives and dispatches the letter of terminal device transmission
Breath realizes communication.
Processor 110 is the control centre of server 100, utilizes various interfaces and the entire server 100 of connection
Various pieces by running or execute the software program/or module that are stored in memory 130, and are called and are stored in storage
Data in device 130, the various functions and processing data of execute server 100.Optionally, processor 110 may include one
Or multiple processing units.
Memory 130 can be used for storing software program and module, and processor 110 is stored in memory 130 by operation
Software program and module, thereby executing various function application and data processing.Memory 130 can mainly include storage journey
Sequence area and storage data area, wherein storing program area can application program needed for storage program area, at least one function etc.；
Storage data area can store the data etc. created according to business processing.In addition, memory 130 may include high random access
Memory, can also include nonvolatile memory, a for example, at least disk memory, flush memory device or other are volatile
Property solid-state memory.
It should be noted that above-mentioned structure shown in FIG. 1 is only a kind of example, it is not limited in the embodiment of the present invention.
Sign mentioned in the embodiment of the present invention is pleural indentation sign, without limitation, is only example effect.
Based on foregoing description, Fig. 2 illustratively shows a kind of process of Lung neoplasm analysis provided in an embodiment of the present invention,
The device that the process can be analyzed by Lung neoplasm executes.
As shown in Fig. 2, the process specifically includes:
Step 201, the three-dimensional of Lung neoplasm in the Lung neoplasm image and the Lung neoplasm image of the lung of patient is obtained to sit
It marks, the segmentation image of the segmentation image of the lung, the Lung neoplasm.
Lung neoplasm image is 3-D image.The three-dimensional coordinate of Lung neoplasm can be the three-dimensional coordinate (ratio of the point in Lung neoplasm
Such as the three-dimensional coordinate of Lung neoplasm central point), it is also possible to the three-dimensional coordinate of the point on Lung neoplasm surface.Lung neoplasm image can be meter
Calculate body layer photography (Computed Tomography, abbreviation CT) image, magnetic resonance imaging (Magnetic Resonance
Imaging, abbreviation MRI) image etc., for clearer description Lung neoplasm image, Fig. 3 illustrates a patient's
Lung's CT images, the sign of the Lung neoplasm in the CT images are pleural indentation sign.The segmentation image of lung can be lung CT
The segmentation image of the image of binaryzation, Lung neoplasm can be the image of the binaryzation of Lung neoplasm.
Step 202, determine to include the Lung neoplasm from the Lung neoplasm image according to the three-dimensional coordinate of the Lung neoplasm
ROI.
Specifically, pre-determined distance can be radiated out centered on the three-dimensional coordinate of Lung neoplasm, determine to include the lung knot
The pixel cube of section, wherein the pre-determined distance be Lung neoplasm radius presupposition multiple, such as Lung neoplasm radius 1.25
Times.Then this pixel cube is intercepted, and interpolation zooms to certain size.Later again to each picture in the pixel cube
Element adds a spatial information channel, exports ROI, spatial information channel is between pixel cube and the three-dimensional coordinate of Lung neoplasm
Distance.For example, it can be centered on the three-dimensional coordinate of Lung neoplasm here, extend L picture to three reference axis all directions
Element, so that it may choose the pixel cube of a 2L*2L*2L size.As shown in fig. 4 a, exemplary for one kind of lung CT image
Schematic diagram.There are Lung neoplasm A in the lung CT image, and the centre coordinate of Lung neoplasm A is (x0, y0, z0), the radius of Lung neoplasm A
It, then can be according to centre coordinate (x for r0, y0, z0) centered on point, it is as shown in Figure 4 b, acquired with 2 times of radius r for side length
Region (square) be the corresponding ROI of Lung neoplasm A.
In view of in actual conditions, the size of each Lung neoplasm is not consistent, and each Lung neoplasm may be irregular component
Shape.Based on the above situation, in the embodiment of the present invention, can centered on the centre coordinate of Lung neoplasm point, with Lung neoplasm
2.5 times, 3 times of major diameter are side length, and using obtained region as ROI.It so, it is possible to guarantee obtained Lung neoplasm image
It can include the region where entire Lung neoplasm, avoid the parts of images for omitting Lung neoplasm.
It should be noted that ROI can be various shapes, cube shaped Lung neoplasm image as described above is corresponding
Region is only a kind of example, and in other possible examples, ROI may be sphere or other shapes.
In other possible implementations for extracting ROI, can also according to position of the Lung neoplasm in lung CT, according to
The profile of Lung neoplasm, extraction obtain ROI, specifically without limitation.
Further, after obtaining ROI, ROI can be subjected to data enhancing, to expand default feature extraction mind
Training sample amount through network model, and then improve the accuracy of default feature extraction neural network model.The embodiment of the present invention
In, there are many modes of data enhancing, for example, can carry out data enhancing by the way of Random Level mirror image to ROI；Or
Person can also carry out data enhancing by the way of the translation of random upper and lower, left and right to ROI；Alternatively, can also be used to ROI
The mode of Random-Rotation carries out data enhancing；Alternatively, can be, using data enhancing is carried out in a manner of scaling at random, to have to ROI
Body is without limitation.
Step 203, by the ROI of the Lung neoplasm, the three-dimensional coordinate of the Lung neoplasm, the segmentation image of the lung and institute
The segmentation image for stating Lung neoplasm is input to default feature extraction neural network model and carries out feature extraction, obtains the Lung neoplasm shadow
The feature vector of picture.
Default feature extraction neural network model can be through marked Lung neoplasm image, marked Lung neoplasm
Three-dimensional coordinate, the segmentation image of marked lung, marked Lung neoplasm segmentation image be trained.Wherein,
Default feature extraction neural network model can be shallow-layer neural network model, i.e., the default feature extraction neural network can wrap
N number of convolution module is included, and, N is less than or equal to first threshold.Those skilled in the art can rule of thumb set with actual conditions
Determine the specific value of first threshold, herein without limitation.
In order to which according to default feature extraction neural network model referred to above is clearly described, Fig. 5 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
Taking neural network model may include three convolution modules.As shown in figure 5, three convolution modules are respectively the first convolution module
501, the second convolution module 502 and third convolution module 503；Wherein, each convolution module may include convolutional layer and convolution again
Normalization (Batch Normalization, BN) layer of layer connection, the activation primitive layer that is connect with BN layers and with activate letter
Pooling the layers of max of several layers of connection, the first convolution module 501 as shown in Figure 5 include the first convolutional layer the 5011, the first BN layers
5012, the first activation primitive layer 5013 and the first max pooling layer 5014, the second convolution module 502 include the second convolutional layer
5021, the 2nd BN layer 5022, the second activation primitive layer 5023 and the 2nd max pooling layer 5024, third convolution module 503 are wrapped
Include third convolutional layer 5031, the 3rd BN layer 5032, third activation primitive layer 5033 and the 3rd max pooling layer 5034.
It should be understood that activation primitive shown in (1) Fig. 5 can be a plurality of types of activation primitives, for example, can be with
For line rectification function (Rectified Linear Unit, ReLU), specifically without limitation；(2) each volume shown in Fig. 5
The feature channel numerical value that the convolution kernel size of lamination, pooling layers of max of convolution kernel size, each convolution module are extracted can be with
It rule of thumb sets for those skilled in the art and adjusts with actual conditions, specifically without limitation；(3) due to the present invention
The image inputted in embodiment is 3-D image, and therefore, the default feature extraction neural network model in the embodiment of the present invention can
Think (3Dimensions, 3D) convolutional neural networks, correspondingly, the corresponding convolution kernel size of 3D convolutional neural networks can be
M*m*m, wherein m is the integer more than or equal to 1.
The method of determination of the corresponding feature vector of patient as described above is only a kind of possible implementation, other
In possible implementation, the corresponding feature vector of patient can also be determined otherwise, specifically without limitation.
By the way that the segmentation image of the segmentation image of the three-dimensional coordinate of the ROI of Lung neoplasm, Lung neoplasm, lung and Lung neoplasm is defeated
Enter to default feature extraction neural network model and carry out feature extraction, obtains the feature vector of Lung neoplasm image.
Step 204, the feature vector of the Lung neoplasm image is input to default Classification Neural model, obtained described
The Lung neoplasm of the patient of default Classification Neural model output analyzes result.
Fig. 6 illustrates a kind of structural representation of default Classification Neural model provided in an embodiment of the present invention
Figure, as shown in fig. 6, the default Classification Neural model includes the first full articulamentum 601, the second complete 602 and of articulamentum
Sigmoid layer 603.The first full articulamentum 601, the second full articulamentum can be passed sequentially through to the corresponding feature vector of patient diagnosed
After 602 are calculated, then output category result after being classified by sigmoid layer 603, to obtain the Lung neoplasm that patient suffers from
Analysis result.
In order to clearly describe default Classification Neural described above to the analytic process of Lung neoplasm, below with
For analyzing the sign of Lung neoplasm, it is specifically described.Wherein, presetting Classification Neural model can be default sign
As Classification Neural model, also, preset the Lung neoplasm that the patient that sign Classification Neural model exports suffers from
Analysis result is the sign for the Lung neoplasm that the patient suffers from.In the embodiment of the present invention, sign Classification Neural model is preset
It is to be trained by the feature vector of the Lung neoplasm image to multiple patients and patient with the known results of Lung neoplasm
It arrives.
Further, the sign of Lung neoplasm may include multiple types, and the embodiment of the present invention is mainly that pleural indentation sign is
Example is described.
It as shown in table 1, is the feature vector of multiple patients and a kind of example of the sign of the Lung neoplasm suffered from.Patient's 1
Feature vector is X1, the sign of the Lung neoplasm suffered from be spicule sign as；The feature vector of patient 2 is X2, the sign of the Lung neoplasm suffered from
As being other；The feature vector of patient 3 is X3, the sign of the Lung neoplasm suffered from is nothing.
A kind of table 1: example of the sign of the feature vector and Lung neoplasm suffered from of the Lung neoplasm image of multiple patients
| Number | Feature vector | The sign of Lung neoplasm |
| Patient 1 | X1 | Pleural indentation sign |
| Patient 2 | X2 | It is other |
| Patient 3 | X3 | Nothing |
| …… | …… | …… |
Further, the feature vector of the Lung neoplasm image of multiple patients shown in table 1 and each patient are suffered from
The sign of Lung neoplasm input in default sign Classification Neural model, can determine default sign Classification Neural model
Parameter.Specifically, the feature vector of the Lung neoplasm image of multiple patients can be first input to initial sign classification mind
Through network model, the prediction sign for the Lung neoplasm that each patient suffers from is obtained, the Lung neoplasm then suffered from according to each patient
The practical sign for the Lung neoplasm that prediction sign and each patient suffer from carries out reverse train, generates default sign classification nerve
Network model.
It should be noted that in the default sign Classification Neural model of training, Lung neoplasm that multiple patients suffer from
Practical sign can be determined by doctor.
For example, shown in the table 1 for patient 1, by the corresponding feature vector, X of patient 11Input default sign point
In neural network, by propagated forward, the result vector of available one 3 dimensionWherein, y1For Pleural indentation
Levy corresponding confidence level；y2For the corresponding confidence level of other signs；y3For the corresponding confidence level of no sign.Classified according to the sign
The result vector of neural networkObtain the corresponding prediction sign of patient 1；Further, if the corresponding prediction of patient 1
Sign is pleural indentation sign, and the corresponding practical sign of patient 1 is other signs according to the content shown in table 1, sign default in this way
As Classification Neural model prediction result and actual result between there is error, i.e., loss (loss) functional value.In turn,
Back-propagation algorithm can be used, according to the algorithm edge stochastic gradient descent (Stochastic gradient descent, SGD)
The direction of loss (loss) functional value decline adjust the parameter of default sign Classification Neural model.In this way, passing through prediction
Comparison between sign and practical sign can accurately adjust the parameter of default sign Classification Neural model, improve life
At default sign Classification Neural model accuracy.
In the embodiment of the present invention, the method for determination of the sign for the Lung neoplasm that patient suffers from can there are many, in an example,
The corresponding feature vector of the patient can be inputted to default Classification Neural model, obtain that multiple default signs are corresponding to set
Reliability, and then can be using the sign for the Lung neoplasm that the highest default sign of confidence level suffers from as patient.
In another example, it is contemplated that there may be a variety of signs for the same Lung neoplasm, for example, the lung knot that certain patient suffers from
The sign of section can be pleural indentation sign and other signs.In such a case, it is possible to first that the corresponding feature vector of patient is defeated
Enter default Classification Neural model, obtains the corresponding confidence level of multiple default signs, and then sign is corresponding to set for presetting
Reliability suffers from if the corresponding confidence level of the default sign is greater than preset threshold using the default sign as the patient
Lung neoplasm sign.
For example, if the corresponding feature vector of patient is X1', by the corresponding feature vector, X of the patient1' the default sign of input
After in classification Model of Neural Network, obtained result vectorWherein, y1' it is the corresponding confidence of pleural indentation sign
Degree；y2' it is the corresponding confidence level of other signs；y3' for the corresponding confidence level of no sign.It is learnt if being computed, y1' be greater than and preset
Threshold value, and y2' and y3' being respectively less than preset threshold, then the sign for the Lung neoplasm that patient suffers from is pleural indentation sign.
In this way, the embodiment of the present invention using default feature extraction neural network model to the ROI of Lung neoplasm, Lung neoplasm three
The segmentation Extraction of Image feature vector for dividing image and Lung neoplasm of coordinate, lung is tieed up, and is come in conjunction with default Classification Neural
Lung neoplasm is analyzed, since model above is by being trained to mass data, so that passing through mould
The result that type obtains is relatively reasonable, and has certain scientific basis.It, can for the mode of traditional diagnosis
Error Diagnostics rate caused by reducing because of doctor's level difference, to improve the accuracy of Lung neoplasm analysis.
Based on the same technical idea, Fig. 7 illustratively shows a kind of Lung neoplasm analysis provided in an embodiment of the present invention
Device 700, the device 700 can execute Lung neoplasm analysis process.
As shown in fig. 7, the device specifically includes:
Acquiring unit 701, Lung neoplasm in the Lung neoplasm image and the Lung neoplasm image for obtaining the lung of patient
Three-dimensional coordinate, the lung segmentation image, the Lung neoplasm segmentation image；
Determination unit 702 determines to include institute for the three-dimensional coordinate according to the Lung neoplasm from the Lung neoplasm image
State the ROI of Lung neoplasm；
Processing unit 703, for by the segmentation of the three-dimensional coordinate, the lung of the ROI of the Lung neoplasm, the Lung neoplasm
The segmentation image of image and the Lung neoplasm is input to default feature extraction neural network model and carries out feature extraction, obtains described
The feature vector of Lung neoplasm image；The default feature extraction neural network model be by marked Lung neoplasm image,
The segmentation image progress of the three-dimensional coordinate of marked Lung neoplasm, the segmentation image of marked lung, marked Lung neoplasm
What training obtained；And the feature vector of the Lung neoplasm image is input to default Classification Neural model, it obtains described
The Lung neoplasm of the patient of default Classification Neural model output analyzes result；Wherein, the default Classification Neural
Model is to be instructed by the feature vector of the Lung neoplasm image to multiple patients and patient with the known results of Lung neoplasm
It gets.
Optionally, the default Classification Neural model is default sign Classification Neural model；The default sign
As Classification Neural model be by the sign of the Lung neoplasm that the corresponding feature vector of multiple patients, each patient are suffered from into
Row training obtains；
The Lung neoplasm analysis result of the patient of the default Classification Neural model output is that the patient suffers from
Lung neoplasm sign.
Optionally, the processing unit 703 is specifically used for:
The corresponding feature vector of the multiple patient is input to initial sign Classification Neural model, is obtained each
The prediction sign for the Lung neoplasm that patient suffers from；
The reality for the Lung neoplasm that the prediction sign of the Lung neoplasm suffered from according to each patient and each patient suffer from
Sign carries out reverse train, generates the default sign Classification Neural model.
Optionally, the processing unit 703 is specifically used for:
The corresponding feature vector of the patient is inputted into default Classification Neural model, obtains that default sign is corresponding sets
Reliability；
If the corresponding confidence level of the default sign is greater than preset threshold, suffer from using the default sign as the patient
The sign of some Lung neoplasms.
Optionally, the sign is pleural indentation sign.
Based on the same technical idea, the embodiment of the invention also provides a kind of calculating equipment, comprising:
Memory, for storing program instruction；
Processor executes above-mentioned lung knot according to the program of acquisition for calling the program instruction stored in the memory
The method for saving analysis.
Based on the same technical idea, the embodiment of the invention also provides a kind of computer-readable non-volatile memories to be situated between
Matter, including computer-readable instruction, when computer is read and executes the computer-readable instruction, so that computer executes
The method for stating Lung neoplasm analysis.
Based on the same technical idea, the embodiment of the invention provides a kind of equipment of Lung neoplasm analysis, as shown in figure 8,
It is unlimited in the embodiment of the present invention including at least one processor 801, and the memory 802 being connect at least one processor
Determine the specific connection medium between processor 801 and memory 802, passes through between processor 801 and memory 802 in Fig. 8 total
For line connection.Bus can be divided into address bus, data/address bus, control bus etc..
In embodiments of the present invention, memory 802 is stored with the instruction that can be executed by least one processor 801, at least
The instruction that one processor 801 is stored by executing memory 802 can be executed and be wrapped in the method for Lung neoplasm analysis above-mentioned
The step of including.
Wherein, processor 801 is the control centre of the equipment of Lung neoplasm analysis, can use various interfaces and connection
The various pieces of the equipment of Lung neoplasm analysis are stored by running or executing the instruction being stored in memory 802 and call
Data in memory 802, to realize that Lung neoplasm is analyzed.Optionally, processor 801 may include that one or more processing are single
Member, processor 801 can integrate application processor and modem processor, wherein the main processing operation system of application processor,
User interface and application program etc., modem processor mainly handle wireless communication.It is understood that above-mentioned modulation /demodulation
Processor can not also be integrated into processor 801.In some embodiments, processor 801 and memory 802 can be same
It is realized on chip, in some embodiments, they can also be realized respectively on independent chip.
Processor 801 can be general processor, such as central processing unit (CPU), digital signal processor, dedicated integrated
Circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array or other can
Perhaps transistor logic, discrete hardware components may be implemented or execute present invention implementation for programmed logic device, discrete gate
Each method, step and logic diagram disclosed in example.General processor can be microprocessor or any conventional processor
Deng.The step of method disclosed in the embodiment analyzed in conjunction with Lung neoplasm, can be embodied directly in hardware processor and execute completion,
Or in processor hardware and software module combination execute completion.
Memory 802 is used as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software journey
Sequence, non-volatile computer executable program and module.Memory 802 may include the storage medium of at least one type,
It such as may include flash memory, hard disk, multimedia card, card-type memory, random access storage device (Random Access
Memory, RAM), static random-access memory (Static Random Access Memory, SRAM), may be programmed read-only deposit
Reservoir (Programmable Read Only Memory, PROM), read-only memory (Read Only Memory, ROM), band
Electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory,
EEPROM), magnetic storage, disk, CD etc..Memory 902 can be used for carrying or storing have instruction or data
The desired program code of structure type and can by any other medium of computer access, but not limited to this.The present invention is real
Applying the memory 802 in example can also be circuit or other devices that arbitrarily can be realized store function, for storing program
Instruction and/or data.
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 (12)
1. a kind of method of Lung neoplasm analysis, which is characterized in that this method comprises:
Obtain the three-dimensional coordinate of Lung neoplasm in the Lung neoplasm image and the Lung neoplasm image of the lung of patient, the lung
Divide the segmentation image of image, the Lung neoplasm；
The area-of-interest comprising the Lung neoplasm is determined from the Lung neoplasm image according to the three-dimensional coordinate of the Lung neoplasm
ROI；
By dividing for the ROI of the Lung neoplasm, the three-dimensional coordinate of the Lung neoplasm, the segmentation image of the lung and the Lung neoplasm
Cut image and be input to default feature extraction neural network model and carry out feature extraction, obtain the feature of the Lung neoplasm image to
Amount；The default feature extraction neural network model is by three to marked Lung neoplasm image, marked Lung neoplasm
Dimension coordinate, the segmentation image of marked lung, marked Lung neoplasm segmentation image be trained；
The feature vector of the Lung neoplasm image is input to default Classification Neural model, obtains the default classification nerve
The Lung neoplasm of the patient of network model output analyzes result；Wherein, the default Classification Neural model is by right
What the feature vector of the Lung neoplasm image of multiple patients and patient were trained with the known results of Lung neoplasm.
2. the method according to claim 1, wherein the default Classification Neural model is default sign point
Connectionist model；The default sign Classification Neural model is by the corresponding feature vector of multiple patients, every
What the sign for the Lung neoplasm that a patient suffers from was trained；
The Lung neoplasm analysis result of the patient of the default Classification Neural model output is the lung that the patient suffers from
The sign of tubercle.
3. according to the method described in claim 2, it is characterized in that, the default Classification Neural model is by multiple
What the feature vector of the Lung neoplasm image of patient and patient were trained with the known sign result of Lung neoplasm, packet
It includes:
The corresponding feature vector of the multiple patient is input to initial sign Classification Neural model, obtains each patient
The prediction sign of the Lung neoplasm suffered from；
The practical sign for the Lung neoplasm that the prediction sign of the Lung neoplasm suffered from according to each patient and each patient suffer from,
Reverse train is carried out, the default sign Classification Neural model is generated.
4. according to the method described in claim 2, it is characterized in that, described input the corresponding feature vector of the patient is preset
Classification Neural model, and obtain point for the Lung neoplasm that the patient that the default Classification Neural model exports suffers from
Analyse result, comprising:
The corresponding feature vector of the patient is inputted into default Classification Neural model, obtains the default corresponding confidence of sign
Degree；
If the corresponding confidence level of the default sign is greater than preset threshold, the default sign is suffered from as the patient
The sign of Lung neoplasm.
5. method according to claim 1 to 4, which is characterized in that the sign is pleural indentation sign.
6. a kind of device of Lung neoplasm analysis characterized by comprising
Acquiring unit, the three-dimensional of Lung neoplasm is sat in the Lung neoplasm image and the Lung neoplasm image for obtaining the lung of patient
It marks, the segmentation image of the segmentation image of the lung, the Lung neoplasm；
Determination unit determines to include the Lung neoplasm for the three-dimensional coordinate according to the Lung neoplasm from the Lung neoplasm image
Region of interest ROI；
Processing unit, for by the segmentation image of the three-dimensional coordinate of the ROI of the Lung neoplasm, the Lung neoplasm, the lung and
The segmentation image of the Lung neoplasm is input to default feature extraction neural network model and carries out feature extraction, obtains the Lung neoplasm
The feature vector of image；The default feature extraction neural network model is by marked Lung neoplasm image, marked
The three-dimensional coordinate of Lung neoplasm, the segmentation image of marked lung, marked Lung neoplasm segmentation image be trained
It arrives；And the feature vector of the Lung neoplasm image is input to default Classification Neural model, obtain described default point
The Lung neoplasm of the patient of Connectionist model output analyzes result；Wherein, the default Classification Neural model is
It is trained to obtain with the known results of Lung neoplasm by the feature vector of the Lung neoplasm image to multiple patients and patient
's.
7. device according to claim 6, which is characterized in that the default Classification Neural model is default sign point
Connectionist model；The default sign Classification Neural model is by the corresponding feature vector of multiple patients, every
What the sign for the Lung neoplasm that a patient suffers from was trained；
The Lung neoplasm analysis result of the patient of the default Classification Neural model output is the lung that the patient suffers from
The sign of tubercle.
8. device according to claim 7, which is characterized in that the processing unit is specifically used for:
The corresponding feature vector of the multiple patient is input to initial sign Classification Neural model, obtains each patient
The prediction sign of the Lung neoplasm suffered from；
The practical sign for the Lung neoplasm that the prediction sign of the Lung neoplasm suffered from according to each patient and each patient suffer from,
Reverse train is carried out, the default sign Classification Neural model is generated.
9. device according to claim 7, which is characterized in that the processing unit is specifically used for:
The corresponding feature vector of the patient is inputted into default Classification Neural model, obtains the default corresponding confidence of sign
Degree；
If the corresponding confidence level of the default sign is greater than preset threshold, the default sign is suffered from as the patient
The sign of Lung neoplasm.
10. device according to any one of claims 6 to 9, which is characterized in that the sign is pleural indentation sign.
11. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer can
It executes instruction, the computer executable instructions are for executing the computer as described in any one of claims 1 to 5
Method.
12. a kind of calculating equipment of image recognition characterized by comprising
Memory, for storing program instruction；
Processor, for calling the program instruction stored in the memory, according to acquisition program execute as claim 1 to
Method described in any one of 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811570841.0A CN109523546A (en) | 2018-12-21 | 2018-12-21 | A kind of method and device of Lung neoplasm analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811570841.0A CN109523546A (en) | 2018-12-21 | 2018-12-21 | A kind of method and device of Lung neoplasm analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109523546A true CN109523546A (en) | 2019-03-26 |
Family
ID=65795989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811570841.0A Pending CN109523546A (en) | 2018-12-21 | 2018-12-21 | A kind of method and device of Lung neoplasm analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109523546A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110349653A (en) * | 2019-07-15 | 2019-10-18 | 杭州依图医疗技术有限公司 | Display methods, equipment and the storage medium of image analysing computer data |
| CN110400295A (en) * | 2019-07-19 | 2019-11-01 | 杭州依图医疗技术有限公司 | A kind of method and device that Lung neoplasm relative position determines |
| CN110826557A (en) * | 2019-10-25 | 2020-02-21 | 杭州依图医疗技术有限公司 | Method and device for detecting fracture |
| CN110889836A (en) * | 2019-11-22 | 2020-03-17 | 中国人民解放军第四军医大学 | Image data analysis method and device, terminal equipment and storage medium |
| CN111476774A (en) * | 2020-04-07 | 2020-07-31 | 广州柏视医疗科技有限公司 | Intelligent sign recognition device based on novel coronavirus pneumonia CT detection |
| CN111784700A (en) * | 2019-04-04 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Lung lobe segmentation, model training, model construction and segmentation method, system and equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004074982A2 (en) * | 2003-02-14 | 2004-09-02 | The University Of Chicago | Method of training massive training artificial neural networks (mtann) for the detection of abnormalities in medical images |
| CN108615237A (en) * | 2018-05-08 | 2018-10-02 | 上海商汤智能科技有限公司 | A kind of method for processing lung images and image processing equipment |
| CN108648179A (en) * | 2018-04-17 | 2018-10-12 | 杭州依图医疗技术有限公司 | A kind of method and device of analysis Lung neoplasm |
| CN108648172A (en) * | 2018-03-30 | 2018-10-12 | 四川元匠科技有限公司 | A kind of CT figure Lung neoplasm detecting systems based on 3D-Unet |
-
2018
- 2018-12-21 CN CN201811570841.0A patent/CN109523546A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2004074982A2 (en) * | 2003-02-14 | 2004-09-02 | The University Of Chicago | Method of training massive training artificial neural networks (mtann) for the detection of abnormalities in medical images |
| CN108648172A (en) * | 2018-03-30 | 2018-10-12 | 四川元匠科技有限公司 | A kind of CT figure Lung neoplasm detecting systems based on 3D-Unet |
| CN108648179A (en) * | 2018-04-17 | 2018-10-12 | 杭州依图医疗技术有限公司 | A kind of method and device of analysis Lung neoplasm |
| CN108615237A (en) * | 2018-05-08 | 2018-10-02 | 上海商汤智能科技有限公司 | A kind of method for processing lung images and image processing equipment |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111784700A (en) * | 2019-04-04 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Lung lobe segmentation, model training, model construction and segmentation method, system and equipment |
| CN111784700B (en) * | 2019-04-04 | 2022-07-22 | 阿里巴巴集团控股有限公司 | Lung lobe segmentation, model training, model construction and segmentation method, system and equipment |
| CN110349653A (en) * | 2019-07-15 | 2019-10-18 | 杭州依图医疗技术有限公司 | Display methods, equipment and the storage medium of image analysing computer data |
| CN110400295A (en) * | 2019-07-19 | 2019-11-01 | 杭州依图医疗技术有限公司 | A kind of method and device that Lung neoplasm relative position determines |
| CN110826557A (en) * | 2019-10-25 | 2020-02-21 | 杭州依图医疗技术有限公司 | Method and device for detecting fracture |
| CN110889836A (en) * | 2019-11-22 | 2020-03-17 | 中国人民解放军第四军医大学 | Image data analysis method and device, terminal equipment and storage medium |
| CN111476774A (en) * | 2020-04-07 | 2020-07-31 | 广州柏视医疗科技有限公司 | Intelligent sign recognition device based on novel coronavirus pneumonia CT detection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109523546A (en) | A kind of method and device of Lung neoplasm analysis | |
| US11200424B2 (en) | Space-time memory network for locating target object in video content | |
| CN111429460B (en) | Image segmentation method, image segmentation model training method, device and storage medium | |
| CN109685819B (en) | Three-dimensional medical image segmentation method based on feature enhancement | |
| CN107784282A (en) | The recognition methods of object properties, apparatus and system | |
| CN109447963A (en) | A kind of method and device of brain phantom identification | |
| CN108986115B (en) | Medical image segmentation method and device and intelligent terminal | |
| KR101603019B1 (en) | Image processing apparatus, image processing method and computer readable medium | |
| CN109711315A (en) | A kind of method and device of Lung neoplasm analysis | |
| CN108648179A (en) | A kind of method and device of analysis Lung neoplasm | |
| WO2021073120A1 (en) | Method and device for marking lung area shadows in medical image, server, and storage medium | |
| An et al. | Medical image segmentation algorithm based on multilayer boundary perception-self attention deep learning model | |
| CN113077479A (en) | Automatic segmentation method, system, terminal and medium for acute ischemic stroke focus | |
| CN112183541B (en) | Contour extraction method and device, electronic equipment and storage medium | |
| US11854193B2 (en) | Validity evaluation device for cancer region detection | |
| US11769594B2 (en) | Deep learning model learning device and method for cancer region | |
| CN113838067B (en) | Method and device for segmenting lung nodules, computing device and storable medium | |
| CN112102230A (en) | Ultrasonic tangent plane identification method, system, computer equipment and storage medium | |
| Yang et al. | Directional connectivity-based segmentation of medical images | |
| Wei et al. | Attentive boundary aware network for multi-scale skin lesion segmentation with adversarial training | |
| CN110992310A (en) | Method and device for determining partition where mediastinal lymph node is located | |
| CN113160199B (en) | Image recognition method and device, computer equipment and storage medium | |
| CN109801262A (en) | A kind of method and device of Lung neoplasm analysis | |
| CN109801264A (en) | A kind of method and device of Lung neoplasm analysis | |
| CN116664590A (en) | Automatic segmentation method and device based on dynamic contrast enhancement magnetic resonance image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190326 |