CN107423576A - A kind of lung cancer identifying system based on deep neural network - Google Patents
A kind of lung cancer identifying system based on deep neural network Download PDFInfo
- Publication number
- CN107423576A CN107423576A CN201710752237.9A CN201710752237A CN107423576A CN 107423576 A CN107423576 A CN 107423576A CN 201710752237 A CN201710752237 A CN 201710752237A CN 107423576 A CN107423576 A CN 107423576A
- Authority
- CN
- China
- Prior art keywords
- neural network
- lung cancer
- network
- application server
- training
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The present invention proposes a kind of lung cancer identifying system based on deep neural network, it is related to medical image processing technology field, including teller system and neural metwork training system, isolated between teller system and neural metwork training system by gateway, the neural metwork training system includes application server, management system, GPU cluster, medical image database, the teller system is used to provide the user service interface, and it includes web page, business logic processing module and database.The present invention improves image preprocessing and the artificial complexity for participating in carrying out feature extraction in conventional method using the automatic learning characteristic of deep learning, reduce the quality requirement to data set, early found for patient, early treatment has won the time, help can be provided for the diagnosis of doctor, the mistaken diagnosis caused by thinking carelessness is prevented, has in the degree of accuracy and on the training time very big advantage.
Description
Technical field
The present invention relates to medical image processing technology field, especially a kind of lung cancer identification system based on deep neural network
System.
Background technology
Lung cancer is to threaten one of human health and the malignant tumour of most serious of life.Research finds, and if can be early
It was found that will be improved with treatment, 5 annual survival rates of lung cancer patient close to 50%, the analysis of histopathology image is pulmonary cancer diagnosis
Goldstandard.But the cell included in every histopathology picture is all 10,000,000,000 ranks, the only manually lookup to cancer cell
It is big with diagnostic work amount, and easily malfunction.Therefore it is very hot at present that histopathology picture, which carries out automatic detection and analysis,
One research direction of door.Have at present some apply conventional machines study methods, as SVM methods to cancer cell carry out detection and
Classification.The degree of accuracy of these algorithms depends on the feature extracting method based on engineer, such as color to cell, texture, shape
The low-level image features such as shape.
Deep learning is a new field in machine learning research, and its object is to establish, simulate human brain to be divided
The neutral net of study is analysed, the mechanism that it imitates human brain explains data, such as image, sound and text.Deep learning at present
It has been widely used in the fields such as speech recognition, image recognition, natural language processing, has promoted the advance of artificial intelligence, bring
One New Wave of machine learning, by the extensive attention from academia to industrial quarters.
The content of the invention
The present invention provides a kind of lung cancer identifying system based on deep neural network, prevents caused by human negligence by mistake
Examine, for patient early has found, early treatment has won the time.
The present invention specifically adopts the following technical scheme that realization：
A kind of lung cancer identifying system based on deep neural network, including teller system and neural metwork training system
System, is isolated between the teller system and neural metwork training system by gateway, wherein,
The neural metwork training system includes application server, management system, GPU cluster, medical image database, institute
Medical image database is stated to be used to store normal human lung's medical imaging and human lung's medical imaging with lung cancer, institute
State management system and control the GPU cluster, the application server includes neural network model, carries out the identification of lung cancer；
The teller system is used to provide the user service interface, and it includes web page, business logic processing module
And database, user submit the lung's medical imaging picture for needing to diagnose, the service logic by accessing the web page
Processing module is sent to the application server after being used for the picture pretreatment by the submission of user, and the application server utilizes
The neutral net trained is analyzed the picture, and analysis result then is fed back into the Service Processing Module, and will
Analysis result stores in the database, and the Service Processing Module is fed back to user again.
Preferably, the training method of the neural metwork training system is supervised learning or learns without tutor or supervise certainly
Educational inspector practises or has tutor and without tutor's blended learning.
Preferably, the training method of the neural metwork training system is as follows：
Step 1, design neural network model；
Step 2, obtain a large amount of lung's pictures；
Step 3, training neutral net, if reaching required precision, carry out step 4, if not up to required precision, carry out
Step 5；
Step 4, it is supplied to application server；
Step 5, backpropagation, adjust neural network model, after multiple iteration, return to step 3.
Preferably, the neural network model is the feedforward network without feedback or has in the feedforward network or layer of feedback
There is the feedforward network of feedback or a type network that be combined with each other.
A kind of lung cancer identifying system based on deep neural network provided by the invention, its advantage are：Utilize depth
The automatic learning characteristic of degree study improves image preprocessing and the artificial complexity for participating in carrying out feature extraction in conventional method
Property, the quality requirement to data set is reduced, giving model to application server automatically after neural metwork training is completed is
User provides service, and user, which need to only provide medical imaging, can learn diagnostic result, for when patient is early had found, early treatment has been won
Between, help can be provided for the diagnosis of doctor, prevent the mistaken diagnosis caused by thinking carelessness.Compared with traditional cell detection method
Compared with the application is in the degree of accuracy and has very big advantage on the training time.
Brief description of the drawings
Fig. 1 is the theory diagram of lung cancer identifying system of the present invention；
Fig. 2 is the theory diagram of neural metwork training system；
Fig. 3 is the training flow chart of neural metwork training system；
Fig. 4 is neural network learning flow chart；
Fig. 5 is the theory diagram of teller system；
Fig. 6 is the workflow diagram of teller system.
Embodiment
To further illustrate each embodiment, the present invention is provided with accompanying drawing.These accompanying drawings are the invention discloses the one of content
Point, it can coordinate the associated description of specification to explain the operation principles of embodiment mainly to illustrate embodiment.Coordinate ginseng
These contents are examined, those of ordinary skill in the art will be understood that other possible embodiments and advantages of the present invention.In figure
Component be not necessarily to scale, and similar element numbers are conventionally used to indicate similar component.
In conjunction with the drawings and specific embodiments, the present invention is further described.
As shown in figure 1, a kind of lung cancer identifying system based on deep neural network that this implementation provides, including user service
System and neural metwork training system, are isolated between teller system and neural metwork training system by gateway.
Wherein, as shown in Fig. 2 neural metwork training system includes application server, management system, GPU cluster, medical shadow
As database, medical image database is used to store normal human lung's medical imaging and human lung's medical treatment with lung cancer
Image, management system control the GPU cluster, and application server includes the neural network model trained, carries out lung cancer
Identification.As shown in figure 3, specific training method is as follows：
Step 1, design neural network model；
Step 2, obtain a large amount of lung's pictures；
Step 3, training neutral net, if reaching required precision, carry out step 4, if not up to required precision, carry out
Step 5；
Step 4, it is supplied to application server；
Step 5, backpropagation, adjust neural network model, after multiple iteration, return to step 3.
Study is a kind of neutral net most important the characteristics of also most making one notice, and the method learnt is mainly included to lead
Teacher learns or learns without tutor's study or self-supervisory or have tutor and have tutor without tutor's blended learning, the present embodiment selection
Practise, for pattern class attribute to be sorted, oneself knows supervised learning.It needs to prepare a collection of correctly inputoutput data to inciting somebody to action
After input data is loaded into network input, the output of the real response of network with it is desired export (teacher's output) compared with
To error, each connection weight is then changed according to error, network is gone down towards the direction that can correctly respond constantly change, Zhi Daoshi
The output of border response and the difference of desired the output this learning algorithm in an acceptable scope are referred to as error correction algorithms,
The input of corresponding each pattern sample, network output has instructs (supervision) signal to match with its attribute corresponding to one.
And before neutral net is divided into the feedforward network for being free of feedback or has feedback in the feedforward network or layer that have feedback
To network or the type network that be combined with each other.The present embodiment from there is the feedforward network of feedback, by adjust, train make one it is specific
Input cause the output specified, as shown in figure 4, in figure network by constantly compare output and desired value, until net
The output of network and desired value terminate close to training after consistent, usual network under this training method for having supervision containing it is many this
Kind inputoutput pair.
As shown in Figure 5,6, teller system is used to provide the user service interface, and it includes web page, service logic
Processing module and database, user submit the lung's medical imaging picture for needing to diagnose, service logic by accessing web page
Processing module is sent to application server after being used for the picture pretreatment by the submission of user, and application server is utilized and trained
Neutral net the picture is analyzed, analysis result is then fed back into the Service Processing Module, and by analysis result
It is stored in database, Service Processing Module is fed back to user again.
The application improves image preprocessing and artificial ginseng in conventional method using the automatic learning characteristic of deep learning
Complexity with carrying out feature extraction, reduces the quality requirement to data set, automatically will after neural metwork training is completed
Model gives application server and provides the user service, and user, which need to only provide medical imaging, can learn diagnostic result, is patient
Early discovery, early treatment have won the time, can provide help for the diagnosis of doctor, prevent the mistaken diagnosis caused by thinking carelessness.With
Traditional cell detection method compares, and the application is in the degree of accuracy and has very big advantage on the training time.
Although specifically showing and describing the present invention with reference to preferred embodiment, those skilled in the art should be bright
In vain, do not departing from the spirit and scope of the present invention that appended claims are limited, in the form and details can be right
The present invention makes a variety of changes, and is protection scope of the present invention.
Claims (4)
1. a kind of lung cancer identifying system based on deep neural network, it is characterised in that including teller system and nerve net
Network training system, isolated between the teller system and neural metwork training system by gateway, wherein,
The neural metwork training system includes application server, management system, GPU cluster, medical image database, the doctor
Image database is treated to be used to store normal human lung's medical imaging and human lung's medical imaging with lung cancer, the pipe
Reason system controls the GPU cluster, and the application server includes neural network model, carries out the identification of lung cancer；
The teller system is used to provide the user service interface, and it includes web page, business logic processing module sum
According to storehouse, user submits the lung's medical imaging picture for needing to diagnose, the business logic processing by accessing the web page
Module is sent to the application server after being used for the picture pretreatment by the submission of user, and the application server is utilized and instructed
The neutral net perfected is analyzed the picture, and analysis result then is fed back into the Service Processing Module, and will analysis
As a result store in the database, the Service Processing Module is fed back to user again.
A kind of 2. lung cancer identifying system based on deep neural network according to claim 1, it is characterised in that the god
Training method through network training system is supervised learning or without tutor's study or self-supervisory study or has tutor and without tutor
Blended learning.
A kind of 3. lung cancer identifying system based on deep neural network according to claim 1, it is characterised in that the god
Training method through network training system is as follows：
Step 1, design neural network model；
Step 2, obtain a large amount of lung's pictures；
Step 3, training neutral net, if reaching required precision, carry out step 4, if not up to required precision, carry out step
5；
Step 4, it is supplied to application server；
Step 5, backpropagation, adjust neural network model, after multiple iteration, return to step 3.
A kind of 4. lung cancer identifying system based on deep neural network according to claim 3, it is characterised in that the god
For the feedforward network without feedback or there is the feedforward network for having feedback in the feedforward network or layer of feedback or mutually through network model
Mating type network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710752237.9A CN107423576A (en) | 2017-08-28 | 2017-08-28 | A kind of lung cancer identifying system based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710752237.9A CN107423576A (en) | 2017-08-28 | 2017-08-28 | A kind of lung cancer identifying system based on deep neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107423576A true CN107423576A (en) | 2017-12-01 |
Family
ID=60434658
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710752237.9A Pending CN107423576A (en) | 2017-08-28 | 2017-08-28 | A kind of lung cancer identifying system based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107423576A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107967946A (en) * | 2017-12-21 | 2018-04-27 | 武汉大学 | Operating gastroscope real-time auxiliary system and method based on deep learning |
| CN108231189A (en) * | 2017-12-12 | 2018-06-29 | 华南师范大学 | Data processing method and medical diagnosis on disease device based on dual-depth nerve learning network |
| CN108596065A (en) * | 2018-04-13 | 2018-09-28 | 深圳职业技术学院 | One kind is based on deep semantic segmentation marine oil spill detecting system and method |
| CN110400297A (en) * | 2019-07-22 | 2019-11-01 | 中国石油大学(华东) | A kind of stages of lung cancer prediction technique based on deep learning |
| CN112820398A (en) * | 2021-01-22 | 2021-05-18 | 神威超算(北京)科技有限公司 | Artificial intelligence-based lung diagnosis auxiliary system and method |
| CN112823396A (en) * | 2018-10-02 | 2021-05-18 | 翰林大学产学合作团 | Endoscope device and method for diagnosing gastric lesion based on gastric endoscope image obtained in real time |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101517614A (en) * | 2006-09-22 | 2009-08-26 | 皇家飞利浦电子股份有限公司 | Advanced computer-aided diagnosis of lung nodules |
| CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
| CN106529673A (en) * | 2016-11-17 | 2017-03-22 | 北京百度网讯科技有限公司 | Deep learning network training method and device based on artificial intelligence |
| CN107066934A (en) * | 2017-01-23 | 2017-08-18 | 华东交通大学 | Tumor stomach cell image recognition decision maker, method and tumor stomach section identification decision equipment |
-
2017
- 2017-08-28 CN CN201710752237.9A patent/CN107423576A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101517614A (en) * | 2006-09-22 | 2009-08-26 | 皇家飞利浦电子股份有限公司 | Advanced computer-aided diagnosis of lung nodules |
| CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
| CN106529673A (en) * | 2016-11-17 | 2017-03-22 | 北京百度网讯科技有限公司 | Deep learning network training method and device based on artificial intelligence |
| CN107066934A (en) * | 2017-01-23 | 2017-08-18 | 华东交通大学 | Tumor stomach cell image recognition decision maker, method and tumor stomach section identification decision equipment |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108231189A (en) * | 2017-12-12 | 2018-06-29 | 华南师范大学 | Data processing method and medical diagnosis on disease device based on dual-depth nerve learning network |
| CN108231189B (en) * | 2017-12-12 | 2022-12-09 | 华南师范大学 | Data processing method and disease diagnosis device based on double-deep neural learning network |
| CN107967946A (en) * | 2017-12-21 | 2018-04-27 | 武汉大学 | Operating gastroscope real-time auxiliary system and method based on deep learning |
| CN108596065A (en) * | 2018-04-13 | 2018-09-28 | 深圳职业技术学院 | One kind is based on deep semantic segmentation marine oil spill detecting system and method |
| CN112823396A (en) * | 2018-10-02 | 2021-05-18 | 翰林大学产学合作团 | Endoscope device and method for diagnosing gastric lesion based on gastric endoscope image obtained in real time |
| CN110400297A (en) * | 2019-07-22 | 2019-11-01 | 中国石油大学(华东) | A kind of stages of lung cancer prediction technique based on deep learning |
| CN112820398A (en) * | 2021-01-22 | 2021-05-18 | 神威超算(北京)科技有限公司 | Artificial intelligence-based lung diagnosis auxiliary system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107423576A (en) | A kind of lung cancer identifying system based on deep neural network | |
| CN111340819B (en) | Image segmentation method, device and storage medium | |
| Kanevsky et al. | Big data and machine learning in plastic surgery: a new frontier in surgical innovation | |
| Dong et al. | Learning to read chest X-ray images from 16000+ examples using CNN | |
| CN105105743B (en) | Electrocardiogram intelligent processing method based on deep neural network | |
| Tennakoon et al. | Image quality classification for DR screening using convolutional neural networks | |
| WO2019024380A1 (en) | Intelligent traditional chinese medicine diagnosis method, system and traditional chinese medicine system | |
| CN108334899A (en) | Quantify the bone age assessment method of information integration based on hand bone X-ray bone and joint | |
| CN111292839B (en) | Image processing method, image processing device, computer equipment and storage medium | |
| CN106777953A (en) | The analysis method and system of medical image data | |
| DE112018006488T5 (en) | AUTOMATED EXTRACTION OF ECHOCARDIOGRAPHIC MEASUREMENTS FROM MEDICAL IMAGES | |
| CN109102899A (en) | Chinese medicine intelligent assistance system and method based on machine learning and big data | |
| CN110083708A (en) | A kind of medical bodies association analysis method of knowledge based map | |
| Abdel-Karim et al. | How and what can humans learn from being in the loop? Invoking contradiction learning as a measure to make humans smarter | |
| Wu et al. | Generative caption for diabetic retinopathy images | |
| CN111553436A (en) | Training data generation method, model training method and device | |
| Xu et al. | Fully deep learning for slit-lamp photo based nuclear cataract grading | |
| Xia et al. | A multi-modality network for cardiomyopathy death risk prediction with CMR images and clinical information | |
| Gang et al. | Recognition of honeycomb lung in CT images based on improved MobileNet model | |
| Feng et al. | Deep learning for chest radiology: a review | |
| Deshpande et al. | Automated detection of Diabetic Retinopathy using VGG-16 architecture | |
| Thamilselvan | Lung cancer prediction and classification using Adaboost data mining algorithm | |
| CN111046978A (en) | Skin mirror image identification method based on StyleGANs and decision fusion | |
| CN110136113A (en) | A kind of vagina pathology image classification method based on convolutional neural networks | |
| WO2018083853A1 (en) | Visual field sensitivity estimation device, method for controlling visual field sensitivity estimation device, and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| 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: 20171201 |