CN108596868A - Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning - Google Patents
Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning Download PDFInfo
- Publication number
- CN108596868A CN108596868A CN201710619353.3A CN201710619353A CN108596868A CN 108596868 A CN108596868 A CN 108596868A CN 201710619353 A CN201710619353 A CN 201710619353A CN 108596868 A CN108596868 A CN 108596868A
- Authority
- CN
- China
- Prior art keywords
- chest
- lung neoplasm
- image
- neural network
- sample
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/30064—Lung nodule
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses Lung neoplasm recognition methods and systems in a kind of chest DR based on deep learning, this programme to form deep neural network by the sample largely manually marked to train, the depth nerve net identifies Lung neoplasm image feature in chest DR image by autonomous learning pulmonary nodule image feature, and with this.Thus Lung neoplasm identifying schemes can be realized in the chest DR constituted carries out automatic identification to the pulmonary nodule effect characteristics in chest DR image, recognition efficiency is high, accuracy of identification is high, the phenomenon for effectively avoiding missing inspection unidentified, effectively solves the problems of prior art.
Description
Technical field
The present invention relates to image recognition technologys, and in particular to chest DR identification technology.
Background technology
Due to the influence smoked with air pollution, lung cancer is that morbidity and mortality are highest pernicious swollen in the world at present
" number one killer " of tumor and China in Recent Years tumor patient.Middle and advanced stage is belonged to when going to a doctor due to most of patients with lung cancer, has treated
Effect is undesirable, and long term survival rate is very low, and the early stage of lung cancer often shows as Small pulmonary nodule, it is therefore desirable to be done to Small pulmonary nodule
To finding in time and carry out antidiastole.
The existing method that pulmonary nodule image is generated from chest radiograph generally comprises the steps：It is clear to generate
The figure of lung areas；Low frequency variation is removed from clear lung areas, to generate level images；And In Grade image carries out
Grayscale morphologic operation at least once, to generate nodule-bone image.
Pulmonary nodule can be detected with following further step by using nodule-bone image：From chest X
The pulmonary nodule of photo-beat piece.This method comprises the following steps：Identify the potential nodule position in nodule-bone image；To
Each potential nodule position peripheral region in nodule-bone image is detached；And come really using the feature of separated region
Whether fixed potential tubercle is tubercle.
In existing pulmonary nodule image technology, the typically cad technique of feature based engineering.Existing CAD system
It is generally basede on the image of high-resolution thin slice calculating computed tomography imaging (HRCT), it is interested that (identification) is automatically detected on morphology
Position may be detectable state in clinically relevant other structures.When reproducing and showing medical image, CAD system allusion quotation
Mark or identify to type studied position.Label is to cause to the Suspected Area that is marked note that and more into one
Step provides classification or characterization to damage (interested position).
That is, CAD (and/or CADx) system can identify the Microcalcification in chest research
(microcalcifications) tuberculosis or in MSCT is pernicious or benign.CAD system combines dept. of radiology doctor
The professional knowledge of teacher, and generally provide and detect abnormal related second of opinion in medical image, and
Diagnostic recommendations can be reproduced.By supporting the early detection to suspecting the damage for being cancer and classification, CAD system to allow earlier
Intervention can be theoretically that patient generates preferable prognosis.
However existing HRCT is of high cost, popularity rate is low, and practicability is not strong；Such existing method often false positive simultaneously
Rate is very high, and by acceptance level bottom, thus there is no the large-scale uses in actual clinical position.Thereby result in existing rabat
The identification of tubercle is substantially by pure artificial diagosis, this is not only of high cost, efficiency is low, time-consuming；And it is read in manual identified
When piece, it is highly dependent on the personal experience of doctor, there is the phenomenon that failing to pinpoint a disease in diagnosis.
Invention content
For the problems of Lung neoplasm technology in existing chest DR, a kind of efficient and high-precision Lung neoplasm is needed to know
Other scheme.
For this purpose, technical problem to be solved by the invention is to provide Lung neoplasms in a kind of chest DR based on deep learning to know
Other method and system.
In order to solve the above technical problem, the present invention provides the chest DR based on deep learning in Lung neoplasm identification side
Method trains a deep neural network, the depth nerve net to pass through autonomous learning lung by the sample largely manually marked
Portion's tubercle image feature, and Lung neoplasm image feature in chest DR image is identified with this.
Further, the sample includes the positive sample comprising Lung neoplasm patient's rabat and other people rabat samples, is being marked
It filters out candidate samples when note from historical sample data first, then candidate samples is audited.
Further, deep neural network is formed using stochastic gradient descent model training, and accelerates training using GPU.
Further, the deep neural network includes 5 convolutional layers being alternately present, ReLU layers and pond layer and 2
A full articulamentum.
Further, the deep neural network carries out the image of autonomous learning Lung neoplasm by error back propagation model
Feature.
Further, to identify, Lung neoplasm image is special in chest DR image using multithreading service for the deep neural network
Sign.
Solution above-mentioned technical problem, Lung neoplasm identifying system in the chest DR provided by the invention based on deep learning,
Including：
Sample database stores the samples largely manually marked in the sample database；
Neural metwork training module, the neural metwork training module obtain the sample by mark from sample database, and
It is trained to form deep neural network, autonomous learning pulmonary nodule image feature；
Identification module, the identification module call deep neural network to carry out pulmonary nodule image feature to chest DR image
Identification.
Further, in the identifying system further include chest DR acquisition module, the chest DR acquisition module and identification
Module data connects, and chest DR image to be identified is reached identification module.
Further, in the identifying system further include an output module, the output module and identification module data company
It connects, exports the recognition result of identification module.
Thus Lung neoplasm identifying schemes, which can be realized, in the chest DR constituted influences the pulmonary nodule in chest DR image
Feature carries out automatic identification, and recognition efficiency is high, and accuracy of identification is high, the phenomenon for effectively avoiding missing inspection unidentified, effectively solves existing
There is the problems of technology.
On this basis, this programme forms deep neural network by training, is influenced by deep learning pulmonary nodule special
Sign, can be accurately identified the image feature of pulmonary nodule in chest DR image, greatly improve accuracy of identification.
Description of the drawings
It is further illustrated the present invention below in conjunction with the drawings and specific embodiments.
Fig. 1 is neural network structure schematic diagram in present example；
Fig. 2 is Lung neoplasm identifying system composition schematic diagram in the chest DR based on deep learning in present example；
Fig. 3 is the schematic diagram that chest DR image recognition is carried out in present example.
Specific implementation mode
In order to make the technical means, the creative features, the aims and the efficiencies achieved by the present invention be easy to understand, tie below
Conjunction is specifically illustrating, and the present invention is further explained.
For the identification of pulmonary nodule in chest DR, this example approach carries out automatic identification, improves efficiency and effectively avoids
The unidentified phenomenon of missing inspection；Deep learning method, autonomous learning pulmonary nodule image feature is used to be realized with this on this basis
It accurately identifies, effectively improves accuracy of identification.
Specifically, this programme trains a deep neural network by the sample largely manually marked, can independently learn
The image feature of pulmonary nodule is practised, thus identifies pulmonary nodule in chest DR image.
Great amount of samples used herein includes the sample of the positive sample and other people rabats of Lung neoplasm patient's rabat, each sample
This is all labelled with the classification of image.
In this programme when carrying out sample mark, using the mode of man-computer cooperation, i.e., sieved from historical sample data first
Candidate pulmonary nodule sample is selected, then candidate samples are audited by doctor, to be confirmed whether it is pulmonary nodule, to significantly
Improve the efficiency of mark.
The sample so marked to form a deep neural network using stochastic gradient descent model (SGD) to train, together
When in the training process also utilize GPU accelerate training process.
Referring to Fig. 1 which shows the structural schematic diagram of deep neural network in this example.
As seen from the figure, deep neural network is formed more than 7 layers in this example, including 5 convolutional layers being alternately present, ReLU
Layer and pond layer and 2 full articulamentums.
Wherein, convolutional layer is to image (different data window data) and filtering matrix (one group of fixed weight：Due to
The weight of each neuron is fixed, so as to the Filter constant as one) doing inner product, (element multiplication is again one by one
Summation) operation, different Filters can obtain different output datas, such as edge and profile.
The ReLU layers of operation for taking absolute value to the handling result of convolutional layer makes it have the nonlinear spy of network
Sign.
Pond layer, for taking region average or maximum ReLU layers of handling result.
Thus the deep neural network of the layered structure constituted, each layer can be formed by training to respective image spy
The identification function of sign, and layer more rearward can more form more abstract and global feature recognition function.I.e. shallower layer can
Study is simply and the characteristics of image (edges in such as various directions) of part, deeper layer can then acquire more abstract and global spy
Sign.
Deep neural network in this example is specifically based on error back propagation (BP) algorithm and carrys out autonomous learning Lung neoplasm
Image feature.
When the deep neural network carries out autonomous learning training, by from the direction for being input to output when calculating error output
It carries out, and adjusts weights and threshold value and then carried out from the direction for being output to input.When forward-propagating, input signal is made by hidden layer
It generates output signal by nonlinear transformation for output node and is transferred to mistake if reality output is not consistent with desired output
The back-propagation process of difference.Error-duration model is by hidden layer by output error to input layer successively anti-pass, and by error distribution
Give each layer all units, using from the error signal that each layer obtains as the foundation of adjustment each unit weights.
By adjusting input node and hidden node linking intensity and hidden node and output node linking intensity with
And threshold value, so that error is declined along gradient direction, trained by repetition learning, determines network parameter corresponding with minimal error
(weights and threshold value), training stop stopping.
Trained neural network can voluntarily handle output error minimum to the input information of similar sample at this time
By the information of non-linear conversion.
Accordingly, the deep neural network in this example identifies lung in chest DR image using high performance multithreading service
The image feature of portion's tubercle.
For Lung neoplasm identifying schemes in the above-mentioned chest DR based on deep learning, this example furthermore provides can be real
Now it is somebody's turn to do Lung neoplasm identifying system in the chest DR based on Lung neoplasm identifying schemes in the chest DR of deep learning.
Should include mainly sample database 110, god based on Lung neoplasm identifying system 100 in the chest DR of deep learning referring to Fig. 2
Through acquisition module 140 and output module 150 in network training module 120, identification module 130, chest DR.
Wherein, sample database 110, the sample largely manually marked for storing storage, for neural metwork training module 120
Training uses.As needed, the sample in sample database 100 can be according to being adjusted under actual conditions.
Neural metwork training module 120 obtains the sample by mark with 110 data connection of sample database from sample database
This, and be trained and to form deep neural network, which being capable of autonomous learning pulmonary nodule image feature.
Identification module 130, respectively with neural metwork training module 120, rabat acquisition module 140 and output module
150 data connections.The identification module 130 receives the chest DR image to be identified sent in acquisition module 140 in chest DR, calls
Deep neural network carries out pulmonary nodule image feature identification to chest DR image to be identified, and recognition result is reached output mould
Block 150.
Acquisition module 140 in chest DR are connect with identification module and chest DR system data, are obtained from chest DR system
Chest DR image to be identified is taken, and sends it to identification module.
Output module 150, with identification module data connection, the recognition result for exporting identification module.
Thus Lung neoplasm identifying system 100 is before operation in the base chest DR constituted, by neural metwork training module 120 from
Sample database 100 obtains the sample by mark, and is trained the depth to be formed and be capable of autonomous learning pulmonary nodule image feature
Neural network model.
When system operation, chest DR image to be identified is got by acquisition module in chest DR 140, and send it to
Identification module 130；
Identification module 130 after the chest DR image to be identified got, will call deep neural network to rabat into
The identification of row pulmonary nodule image feature, pulmonary nodule image feature of the deep neural network based on autonomous learning, to be identified
Rabat carries out pulmonary nodule image feature identification, and recognition result is reached output module 150；
Output module 150, get identification module 130 transmission identification structure, will externally export, as picture realize,
Text importing, sound report etc..
Thus Lung neoplasm identifying system 100 in the chest DR based on deep learning constituted, in specific application, by big
The sample that manually marks is measured to train a deep neural network, the image feature to pulmonary nodule can be learnt, to right
The chest DR image of input carries out tubercle screening, and provides the probability of suspicious degree.
In practical application, this system can be embedded in PACS system, when called upon, system, which can return to input picture, lung
The probability of portion's tubercle；When probability is more than some threshold value, doctor can be reminded, further to be checked.
Referring to Fig. 3 which shows an application example of Lung neoplasm identifying system in this chest DR.In the application example,
In the equipment for running this system, GPU by the chest DR image tubercle screening of i.e. recognizable input in 0.413 second as a result,
And the probability of suspicious degree is provided, efficiency is very high.
In addition, the case where this system can help doctor effectively to avoid failing to pinpoint a disease in diagnosis in practical application, and effectively improve work.
For example, this system is for the check to artificial diagosis result, to avoid failing to pinpoint a disease in diagnosis, implementation process is as follows.
Radiologist obtains chest DR image and carries out artificial diagosis, report；Report enters system, is not mentioned in report
Tubercle；If this system goes out pulmonary nodule image feature for the chest DR image recognition, doctor's diagosis again is prompted；Finally,
Doctor then changes report as found to truly have pulmonary nodule；Thus it can effectively avoid the case where failing to pinpoint a disease in diagnosis.
Furthermore this system directly carries out chest DR image recognition, ensure that accuracy of identification and efficiency, implementation process are as follows.
System obtains and identifies chest DR image, if identifying pulmonary nodule image feature, is corresponded in chest DR image
Image on prompted, and Auto-writing partial report；The prompt that system provides is seen when doctor's diagosis, can quickly find disease
Stove；Finally the report automatically generated is modified or audited by doctor.
It is last it may be noted that said program, be pure software framework, tangible media can be laid in through program code, such as
Hard disk, floppy disk, disc or any machine-readable (such as smartphone, computer-readable) store media, work as machine
Loading procedure code and execution, if smartphone loads and executes, machine becomes to carry out the device of this programme.
Furthermore said program also can be with form of program codes through some transmission media, such as cable, optical fiber or any
Transmission kenel is transmitted, and when program code is received, loads and executed by machine, such as smartphone, machine becomes to reality
The device of row said program.
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (9)
1. Lung neoplasm recognition methods in a kind of chest DR based on deep learning, which is characterized in that pass through what is largely manually marked
Sample to form deep neural network to train, and the depth nerve net is by autonomous learning pulmonary nodule image feature, and with this
To identify Lung neoplasm image feature in chest DR image.
2. Lung neoplasm recognition methods in chest DR according to claim 1, which is characterized in that the sample includes comprising lung
The positive sample of tubercle patient's rabat and other people rabat samples, candidate sample is filtered out in mark from historical sample data first
This, then candidate samples are audited.
3. Lung neoplasm recognition methods in chest DR according to claim 1, which is characterized in that use stochastic gradient descent mould
Type training forms deep neural network, and accelerates training using GPU.
4. Lung neoplasm recognition methods in chest DR according to claim 1, which is characterized in that the deep neural network packet
Include 5 convolutional layers being alternately present, ReLU layers and pond layer and 2 full articulamentums.
5. Lung neoplasm recognition methods in chest DR according to claim 1, which is characterized in that the deep neural network is logical
Cross the image feature that error back propagation model carries out autonomous learning Lung neoplasm.
6. Lung neoplasm recognition methods in chest DR according to claim 1, which is characterized in that the deep neural network is adopted
Lung neoplasm image feature in chest DR image is identified with multithreading service.
7. Lung neoplasm identifying system in a kind of chest DR based on deep learning, which is characterized in that including：
Sample database stores the samples largely manually marked in the sample database；
Neural metwork training module, the neural metwork training module obtains the sample by mark from sample database, and carries out
Training forms deep neural network, autonomous learning pulmonary nodule image feature；
Identification module, the identification module call deep neural network to carry out pulmonary nodule image feature knowledge to chest DR image
Not.
8. Lung neoplasm identifying system in chest DR according to claim 7, which is characterized in that also wrapped in the identifying system
Chest DR acquisition module, the chest DR acquisition module and identification module data connection are included, chest DR image to be identified is passed
To identification module.
9. Lung neoplasm identifying system in chest DR according to claim 7, which is characterized in that also wrapped in the identifying system
An output module, the output module and identification module data connection are included, the recognition result of identification module is exported.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710619353.3A CN108596868A (en) | 2017-07-26 | 2017-07-26 | Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710619353.3A CN108596868A (en) | 2017-07-26 | 2017-07-26 | Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108596868A true CN108596868A (en) | 2018-09-28 |
Family
ID=63632974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710619353.3A Pending CN108596868A (en) | 2017-07-26 | 2017-07-26 | Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596868A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428375A (en) * | 2019-07-24 | 2019-11-08 | 东软医疗系统股份有限公司 | A kind of processing method and processing device of DR image |
| CN111950584A (en) * | 2020-06-16 | 2020-11-17 | 江西中科九峰智慧医疗科技有限公司 | An intelligent identification method and system for the integrity of parts in an X-ray chest film |
| CN113888519A (en) * | 2021-10-14 | 2022-01-04 | 四川大学华西医院 | A prediction system for predicting the degree of malignancy of solid pulmonary nodules |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001026026A2 (en) * | 1999-10-04 | 2001-04-12 | University Of Florida | Local diagnostic and remote learning neural networks for medical diagnosis |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105574871A (en) * | 2015-12-16 | 2016-05-11 | 深圳市智影医疗科技有限公司 | Segmentation and classification method and system for detecting lung locality lesion in radiation image |
| 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 |
-
2017
- 2017-07-26 CN CN201710619353.3A patent/CN108596868A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001026026A2 (en) * | 1999-10-04 | 2001-04-12 | University Of Florida | Local diagnostic and remote learning neural networks for medical diagnosis |
| CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
| CN105574871A (en) * | 2015-12-16 | 2016-05-11 | 深圳市智影医疗科技有限公司 | Segmentation and classification method and system for detecting lung locality lesion in radiation image |
| 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 |
Non-Patent Citations (2)
| Title |
|---|
| 胡宝民等: "《河北省区域创新系统研究》", 28 February 2006 * |
| 邓力，俞栋著；谢磊译: "《深度学习方法及应用》", 31 December 2015, 机械工业出版社 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428375A (en) * | 2019-07-24 | 2019-11-08 | 东软医疗系统股份有限公司 | A kind of processing method and processing device of DR image |
| CN110428375B (en) * | 2019-07-24 | 2024-03-01 | 东软医疗系统股份有限公司 | DR image processing method and device |
| CN111950584A (en) * | 2020-06-16 | 2020-11-17 | 江西中科九峰智慧医疗科技有限公司 | An intelligent identification method and system for the integrity of parts in an X-ray chest film |
| CN111950584B (en) * | 2020-06-16 | 2024-05-14 | 江西中科九峰智慧医疗科技有限公司 | An intelligent identification method and system for part integrity in chest X-rays |
| CN113888519A (en) * | 2021-10-14 | 2022-01-04 | 四川大学华西医院 | A prediction system for predicting the degree of malignancy of solid pulmonary nodules |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106504232A (en) | A kind of pulmonary nodule automatic testing method based on 3D convolutional neural networks | |
| CN110232383A (en) | A kind of lesion image recognition methods and lesion image identifying system based on deep learning model | |
| CN109272048A (en) | A kind of mode identification method based on depth convolutional neural networks | |
| CN108230311A (en) | A kind of breast cancer detection method and device | |
| CN107729911A (en) | A kind of pulmonary tuberculosis intelligent identification Method and system based on DR | |
| CN107730484A (en) | A kind of abnormal rabat intelligent identification Method and system based on deep learning | |
| US20200134815A1 (en) | System and Method for an Automated Parsing Pipeline for Anatomical Localization and Condition Classification | |
| CN108596198A (en) | A kind of recognition methods of pneumothorax x-ray image and system based on deep learning | |
| JP2018175226A (en) | Medical image classification device, method, and program | |
| CN1934589A (en) | Systems and methods providing automated decision support for medical imaging | |
| JP7612693B2 (en) | Systems and methods for analyzing two-dimensional and three-dimensional image data - Patents.com | |
| CN109124660B (en) | Gastrointestinal stromal tumor postoperative risk detection method and system based on deep learning | |
| Shen et al. | Guest editorial: special issue on imaging-based diagnosis of COVID-19 | |
| CN108596868A (en) | Lung neoplasm recognition methods and system in a kind of chest DR based on deep learning | |
| CN107578405A (en) | A kind of pulmonary nodule automatic testing method based on depth convolutional neural networks | |
| Dovganich et al. | Automatic out-of-distribution detection methods for improving the deep learning classification of pulmonary X-ray images | |
| WO2022016262A1 (en) | Lung ultrasound processing systems and methods | |
| CN114240874A (en) | Bone age assessment method, device and computer-readable storage medium based on deep convolutional neural network and feature fusion | |
| CN113344887A (en) | Interstitial pneumonia assessment method based on deep learning and fuzzy logic | |
| CN107256544A (en) | A kind of prostate cancer image diagnosing method and system based on VCG16 | |
| CN117409016A (en) | Automatic segmentation method for magnetic resonance image | |
| Fauci et al. | A massive lesion detection algorithm in mammography | |
| CN115049879B (en) | Recognition and classification method of plant leaf disease severity and type based on convolutional Swin Transformer | |
| US12068077B2 (en) | Application of artificial intelligence on detecting canine left atrial enlargement on thoracic radiographs | |
| Ho et al. | Human-airway surface mesh smoothing based on graph convolutional neural networks |
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: 20180928 |