CN108986073A - A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame - Google Patents
A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame Download PDFInfo
- Publication number
- CN108986073A CN108986073A CN201810563693.3A CN201810563693A CN108986073A CN 108986073 A CN108986073 A CN 108986073A CN 201810563693 A CN201810563693 A CN 201810563693A CN 108986073 A CN108986073 A CN 108986073A
- Authority
- CN
- China
- Prior art keywords
- network
- false positive
- candidate nodule
- candidate
- nodule
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
- G06T2207/30064—Lung nodule
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (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 CT image pulmonary nodule detection methods based on improved Faster R-CNN frame, include the following steps: the CT image of the chest of 1, acquisition Lung neoplasm patients with symptom, and mark Lung neoplasm position, as training sample set；2 building candidate nodules detect network, detect network with the training sample set training candidate nodule, determine network parameter, obtain candidate nodule detection model；3, building candidate nodule false positive rejects sorter network, rejects sorter network with the training sample set training candidate nodule false positive, obtains candidate nodule false positive and reject sorter network model；4, CT image to be detected is inputted in candidate nodule detection model, obtains the position of candidate nodule；The position input candidate nodule false positive of the candidate nodule is rejected in sorter network model, false positive is rejected, obtains Lung neoplasm testing result.This method is more suitable for the detection of small size Lung neoplasm.
Description
Technical field
The present invention relates to a kind of pulmonary nodule detection methods of CT image, more particularly to one kind to be based on improved Faster R-
The CT image pulmonary nodule detection method of CNN frame, belongs to Lung neoplasm detection technique field.
Background technique
X ray computer tomographic imaging (X-ray Computer Tomography, CT) technology is by carrying out to object
Ray projection measures and obtains the imaging technique of the accurate and lossless cross section dampening information of object, is that current routine is effectively faced
One of bed medical diagnostic tool, provides 3 D human body organ-tissue information abundant for the diagnosis and prevention of clinician, has become
For inspection diagnostic method indispensable in medical imaging field.
Lung cancer is the highest cancer of the death rate in various cancers, and the death rate that wherein male suffers from lung cancer is 13%, Nv Xingshi
19.5%.About 70% patient is just to be diagnosed in advanced lung cancer, and 5 annual survival rates in this case are only big
About 16% or so.However, if capable of being diagnosed to be early stage lung cancer as a result, so 5 years survival rates can reach 70%.
Lung neoplasm is the old model of lung cancer, this is but also the detection of Lung neoplasm is extremely important.CT image has precision high, using wide
General feature, so the CT picture of chest is the conventional means of Lung neoplasm detection.
The explanation of lung CT image is a very challenging task.There is the institutional framework of many complexity in lung,
The gray value of these tissues is close simultaneously, these all make the detection difficult of Lung neoplasm, especially there is the diameter of many Lung neoplasms
Very small, size only has several millimeters, it is easier to missing inspection.Even experienced radiologist is also in more difficult differentiation lung CT
Blood vessel and the lesser Lung neoplasm of diameter, while the judgement of people also suffers from subjective factor and is influenced.
Traditional Lung neoplasm detection generally comprises the steps: (1) pretreatment of CT image, the segmentation of (2) pulmonary parenchyma, (3)
Candidate nodule detection, and (4) candidate nodule false positive are rejected.Wherein CT image pre-processing phase typically only uses geometry
Learn feature and textural characteristics.Stage is divided for pulmonary parenchyma, usually threshold method and morphologic method is combined to be partitioned into
Pulmonary parenchyma.For candidate nodule detection-phase, threshold method, statistics and Matching Model and the side for combining many algorithms are generally used
Method detects candidate nodule.The candidate nodule false positive rejecting stage usually combines different features (feature of gray value, shape feature
And textural characteristics), using based on engineer method, support vector machines (Support Vector Machine, SVM),
Classifier based on linear discriminant analysis (Linear Discriminant Analysis, LDA) is based on decision tree
The feature classifiers such as the classification method of (Decision tree) screen false positive.Traditional pulmonary nodule detection method is to can only examine
Measure big tubercle, and tubercle lesser for size, it occur frequently that missing inspection.
Summary of the invention
Goal of the invention: aiming at the problems existing in the prior art, the present invention provides a kind of CT image Lung neoplasm detection sides
Method, this method are more suitable for the detection of small size Lung neoplasm.
Technical solution: the present invention adopts the following technical scheme:
A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame, comprising steps of
(1) the CT image of the chest of Lung neoplasm patients with symptom is acquired, and marks Lung neoplasm position, as training sample set；
(2) building candidate nodule detects network, detects network with the training sample set training candidate nodule, determines network
Parameter obtains candidate nodule detection model；
(3) building candidate nodule false positive rejects sorter network, with the training sample set training candidate nodule false positive
Sorter network is rejected, candidate nodule false positive is obtained and rejects sorter network model；
(4) CT image to be detected is inputted in candidate nodule detection model, obtains the position of candidate nodule；By the time
It selects the position of tubercle to input candidate nodule false positive to reject in sorter network model, rejects false positive, obtain Lung neoplasm detection knot
Fruit.
The candidate nodule detection network is the fast convolution neural network Faster R-CNN model based on region, described
Faster R-CNN model includes multiple convolutional layers and a warp lamination.
It is VGG-16 structural network that the candidate nodule false positive, which rejects sorter network,.
The training sample synthesizes an image as a sample using three adjacent slices.
The step (3) includes: the Lung neoplasm position that sample mark is concentrated according to training sample, training Lung neoplasm and non-lung
The sorter network of tubercle obtains candidate nodule false positive and rejects sorter network model.
The utility model has the advantages that compared with prior art, the CT figure disclosed by the invention based on improved Faster R-CNN frame
It, can as pulmonary nodule detection method has the advantage that 1, increases warp lamination after the convolutional layer of Faster R-CNN model
The feature that network extracts before playing the role of amplification, is more suitable for the detection of small size Lung neoplasm；2, due to deep learning pair
The learning ability of high-dimensional feature is very strong, and while not influencing susceptibility and false positive, method disclosed by the invention is omitted
Pulmonary parenchyma segmentation step in traditional Lung neoplasm detection, thus the testing process simplified；3, using acquisition candidate nodule and candidate
Tubercle false positive is rejected two stages and is detected, and is conducive to promote detection effect, while two stages of training are most of all
It is carried out in GPU, training speed is also guaranteed.
Detailed description of the invention
Fig. 1 is sample instantiation figure in sample set in the present invention；
Fig. 2 is the structural schematic diagram of candidate nodule detection model in the present invention；
Fig. 3 is that the Lung neoplasm classification that false positive rejects stage-training data in the present invention pre-processes schematic diagram；
Fig. 4 is that the non-Lung neoplasm classification that false positive rejects stage-training data in the present invention pre-processes schematic diagram；
Fig. 5 is detection-phase candidate nodule testing result diagram in the embodiment of the present invention；
Fig. 6 is the result diagram in the embodiment of the present invention after the rejecting of detection-phase false positive.
Specific embodiment
With reference to the accompanying drawings and detailed description, the present invention is furture elucidated.
A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame, comprising steps of
The CT image of the chest of step 1, acquisition Lung neoplasm patients with symptom, and Lung neoplasm position is marked, as training sample
Collection；
Lung's Medical imaging alliance (The Lung Image Database is used in the present embodiment
Consortium, LIDC) disclosed in data set, contain 1018 cases in this data set, at the same each case have it is corresponding
The diagnostic message that provides of exper ienced radiologist (usually three either four radiologists).
When handling the sample in data set, for the CT picture of each patient, first three dimensions are interpolated into simultaneously
1mm×1mm×1mm；Since [- 1200,600] HU is the more appropriate range for observing lung, [- 1200,600] area HU is chosen
Between be mapped to [0,255] pixel；Then three adjacent slices are chosen, the gray value in respective pixel is added, synthesizes one
Open image as a sample, as shown in Figure 1.Each slice is single channel, and natural image is triple channel, is superimposed three layers of slice
Gray value forms triple channel image, the advantage of doing so is that taking full advantage of the information between adjacent three slices.It is corresponding position
The superposition of upper three slices gray value.
Step 2, building candidate nodule detect network, detect network with the training sample set training candidate nodule, determine
Network parameter obtains candidate nodule detection model；
The candidate nodule detection network is the fast convolution neural network Faster R-CNN model based on region, described
Faster R-CNN model includes multiple convolutional layers and a warp lamination.
Assuming that the input picture size of convolutional layer is W × W, convolution kernel having a size of F × F, step-length S, pixel filling
Padding is P, then the size exported after convolution are as follows:
Assuming that W=7, F=3, S=2, P=0, then the size exported are as follows:
For deconvolution, it is also assumed that input picture size is W × W, and convolution kernel is filled out having a size of F × F, step-length S, pixel
The padding filled is P, then the output size after deconvolution are as follows:
N=(W-1) × S+F-2*P
Assuming that W=3, F=3, S=2, P=0.The size so exported are as follows:
N=(3-1) × 2+3-2 × 0=7
It can reduce the resolution ratio of characteristic pattern by convolution it can be seen from the calculating of convolution above and deconvolution, and warp
Product can play the role of up-sampling, being capable of amplification characteristic figure.
As shown in Fig. 2, increased warp lamination can play the role of the feature extracted of network before amplification, it is suitble to small
The detection of size Lung neoplasm.
Step 3, building candidate nodule false positive reject sorter network, with the false sun of the training sample set training candidate nodule
Property reject sorter network, obtain candidate nodule false positive reject sorter network model；
It is VGG-16 structural network that the candidate nodule false positive, which rejects sorter network,.
The Lung neoplasm position of sample mark is concentrated according to training sample, trains the sorter network of Lung neoplasm and non-Lung neoplasm,
It obtains candidate nodule false positive and rejects sorter network model.
In the present embodiment, Lung neoplasm training set processing step are as follows: according to the location information of mark, find out with Lung neoplasm
Position selects 32 × 32 fritter centered on Lung neoplasm, three adjacent slices is synthesized a sample, such as Fig. 3 institute
Show；The processing step of non-Lung neoplasm training set are as follows: the part of (non-Lung neoplasm) except selection labeling position, having a size of 32 × 32
Three adjacent slices are synthesized a sample, as shown in Figure 4 by fritter.Lung neoplasm and non-Lung neoplasm two types are established
Sorter network, as shown in table 1, in table 1 in title column, conv1_1 represents first convolutional layer of first stage, conv1_2 generation
Second convolutional layer of table first stage, and so on.Pooling1 represents the pond layer of first stage, and pooling2 represents second
Stage pond layer, and so on.It is full articulamentum that fc5, which represented for the 5th stage, and it is full articulamentum that fc6, which represented for the 6th stage, with this
Analogize.In parameter setting, what size was represented is the size of convolution kernel, and num represents channel number, and stride represents step-length.Network
The convolution kernel size that middle convolutional layer uses is all 3 × 3, every layer of port number since 32, after each pond according to 2 multiple
Increase, to the last reaches 256；Convolution kernel in the layer of pond is 2 × 2.After training, available false positive rejects classification
Network model.
1 false positive of table rejects network structure
| Title | Type | Parameter setting |
| conv1_1 | convolution | size:3×3,num:32stride:1 |
| conv1_2 | convolution | size:3×3,num:32stride:1 |
| pooling1 | max pooling | size:2×2,stride:2 |
| conv2_1 | convolution | size:3×3,num:64,stride:1 |
| conv2_2 | convolution | size:3×3,num:64,stride:1 |
| pooling2 | max pooling | size:2×2,stride:2 |
| conv3_1 | convolution | size:3×3,num:128,stride:1 |
| conv3_2 | convolution | size:3×3,num:128,stride:1 |
| conv3_3 | convolution | size:3×3,num:128,stride:1 |
| pooling3 | max pooling | size:2×2,stride:2 |
| conv4_1 | convolution | size:3×3,num:256,stride:1 |
| conv4_2 | convolution | size:3×3,num:256,stride:1 |
| conv4_3 | convolution | size:3×3,num:256,stride:1 |
| pooling4 | max pooling | size:2×2,stride:2 |
| fc5 | fully connected | num:1024 |
| fc6 | fully connected | num:1024 |
| fc7 | fully connected | num:2 |
Step 4 inputs CT image to be detected in candidate nodule detection model, obtains the position of candidate nodule；By institute
The position input candidate nodule false positive for stating candidate nodule is rejected in sorter network model, is rejected false positive, is obtained Lung neoplasm inspection
Survey result.
Two CT images to be detected are detected in the present embodiment, obtained candidate nodule position respectively as shown in figure 5,
Wherein Fig. 5-(a) is the candidate Lung neoplasm that two detected are connected with blood vessel；Fig. 5-(b) is that two sizes detecting are smaller
Candidate Lung neoplasm.Rejecting false positive is carried out to candidate Lung neoplasm, obtained final result as shown in fig. 6, wherein Fig. 6-(a) with
Fig. 5-(a) is corresponding, eliminates the false-positive nodule below picture；Fig. 6-(b) is corresponding with Fig. 5-(b), eliminates and is located at figure
False-positive nodule on the left of piece.
The effect of presently disclosed method can be assessed from two angles.First aspect is that observation size is lesser
Whether Lung neoplasm or the Lung neoplasm being attached on blood vessel can correctly detect result.As shown in Fig. 6-(a), this Lung neoplasm is attached
On blood vessel, it is not easy to differentiated with blood vessel, and testing result accurate positioning；As shown in Fig. 6-(b), this Lung neoplasm size is very
It is small, it is easy missing inspection, method disclosed by the invention still can accurately detected.The second aspect is quick compared with conventional method
Sensitivity.
In order to quantitatively verify the validity of published method of the present invention, susceptibility can be compared to evaluate the excellent of distinct methods
It is bad.True positives, true negative, the definition of false positive and false negative are as shown in table 2:
Table 2
According to the definition in table 2, susceptibility are as follows:
Table 3 compares the result that method proposed by the present invention and traditional pulmonary nodule detection method obtain:
Table 3
| Author | Susceptibility (%) |
| Aresta | 57.4 |
| Santos | 80.5 |
| Jacobs | 80.0 |
| Teamoto and Fujita | 80.0 |
| Guo and Li | 80.0 |
| Method disclosed by the invention | 82.3 |
It is the detection of conventional method Lung neoplasm recent years in table 3 as a result, and at the same time using LIDC data set as instruction
Practice collection, the susceptibility that the method for the present invention obtains is higher than the susceptibility of other methods, this has also strongly suggested the method for the present invention
Validity.
As can be seen that the susceptibility of the method for the present invention is superior to other methods from 3 data of table, and due to deep learning
Obvious to high dimensional nonlinear feature learning effect, while not influencing susceptibility, the method for the present invention eliminates pulmonary parenchyma segmentation
Step simplifies the step of Lung neoplasm detects.The enhancing of susceptibility has the computer-aided diagnosis Lung neoplasm of clinical use
Significance.
Claims (5)
1. a kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame, which is characterized in that including such as
Lower step:
(1) the CT image of the chest of Lung neoplasm patients with symptom is acquired, and marks Lung neoplasm position, as training sample set；
(2) building candidate nodule detects network, detects network with the training sample set training candidate nodule, determines that network is joined
Number, obtains candidate nodule detection model；
(3) building candidate nodule false positive rejects sorter network, is rejected with the training sample set training candidate nodule false positive
Sorter network obtains candidate nodule false positive and rejects sorter network model；
(4) CT image to be detected is inputted in candidate nodule detection model, obtains the position of candidate nodule；By the candidate knot
The position input candidate nodule false positive of section is rejected in sorter network model, is rejected false positive, is obtained Lung neoplasm testing result.
2. the CT image pulmonary nodule detection method according to claim 1 based on improved Faster R-CNN frame,
It is characterized in that, the candidate nodule detection network is the fast convolution neural network Faster R-CNN model based on region, institute
Stating Faster R-CNN model includes multiple convolutional layers and a warp lamination.
3. the CT image pulmonary nodule detection method according to claim 1 based on improved Faster R-CNN frame,
It is characterized in that, it is VGG-16 structural network that the candidate nodule false positive, which rejects sorter network,.
4. the CT image pulmonary nodule detection method according to claim 1 based on improved Faster R-CNN frame,
It is characterized in that, the training sample synthesizes an image as a sample using three adjacent slices.
5. the CT image pulmonary nodule detection method according to claim 3 based on improved Faster R-CNN frame,
It is characterized in that, the step (3) includes: the Lung neoplasm position that sample mark is concentrated according to training sample, training Lung neoplasm and non-
The sorter network of Lung neoplasm obtains candidate nodule false positive and rejects sorter network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563693.3A CN108986073A (en) | 2018-06-04 | 2018-06-04 | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563693.3A CN108986073A (en) | 2018-06-04 | 2018-06-04 | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108986073A true CN108986073A (en) | 2018-12-11 |
Family
ID=64539985
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810563693.3A Pending CN108986073A (en) | 2018-06-04 | 2018-06-04 | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108986073A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109805963A (en) * | 2019-03-22 | 2019-05-28 | 深圳开立生物医疗科技股份有限公司 | The determination method and system of one Endometrium parting |
| CN109886307A (en) * | 2019-01-24 | 2019-06-14 | 西安交通大学 | A kind of image detecting method and system based on convolutional neural networks |
| CN109916933A (en) * | 2019-01-04 | 2019-06-21 | 中国人民解放军战略支援部队信息工程大学 | X ray computer tomographic imaging spectra estimation method based on convolutional neural networks |
| CN110136115A (en) * | 2019-05-14 | 2019-08-16 | 重庆大学 | IVOCT image vulnerable plaque detects Approach For Neural Network Ensemble automatically |
| CN110246143A (en) * | 2019-06-14 | 2019-09-17 | 吉林大学第一医院 | Lung CT image assists detection processing device |
| CN110310281A (en) * | 2019-07-10 | 2019-10-08 | 重庆邮电大学 | Lung neoplasm detection and dividing method in a kind of Virtual Medical based on Mask-RCNN deep learning |
| CN110533637A (en) * | 2019-08-02 | 2019-12-03 | 杭州依图医疗技术有限公司 | A kind of method and device of test object |
| CN110827310A (en) * | 2019-11-01 | 2020-02-21 | 北京航空航天大学 | CT image automatic detection method and system |
| 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 |
| CN110838114A (en) * | 2019-11-11 | 2020-02-25 | 苏州锐一仪器科技有限公司 | Pulmonary nodule detection method, device and computer storage medium |
| CN110942446A (en) * | 2019-10-17 | 2020-03-31 | 付冲 | Pulmonary nodule automatic detection method based on CT image |
| CN111311744A (en) * | 2020-04-02 | 2020-06-19 | 佛山市普世医学科技有限责任公司 | Candidate frame filtering, fusion and automatic updating method, system and storage medium for identifying pulmonary nodules |
| CN111798410A (en) * | 2020-06-01 | 2020-10-20 | 深圳市第二人民医院(深圳市转化医学研究院) | Cancer cell pathological grading method, device, equipment and medium based on deep learning model |
| CN112053351A (en) * | 2020-09-08 | 2020-12-08 | 哈尔滨工业大学（威海） | Method for judging benign and malignant pulmonary nodules based on neural network architecture search and attention mechanism |
| CN112529870A (en) * | 2020-12-10 | 2021-03-19 | 重庆大学 | Multi-scale CNNs (CNNs) lung nodule false positive elimination method based on combination of source domain and frequency domain |
| CN112541909A (en) * | 2020-12-22 | 2021-03-23 | 南开大学 | Lung nodule detection method and system based on three-dimensional neural network of slice perception |
| CN112669284A (en) * | 2020-12-29 | 2021-04-16 | 天津大学 | Method for realizing pulmonary nodule detection by generating confrontation network |
| CN113361584A (en) * | 2021-06-01 | 2021-09-07 | 推想医疗科技股份有限公司 | Model training method and device, and pulmonary arterial hypertension measurement method and device |
| CN116012355A (en) * | 2023-02-07 | 2023-04-25 | 重庆大学 | Adaptive false positive lung nodule removing method based on deep learning |
-
2018
- 2018-06-04 CN CN201810563693.3A patent/CN108986073A/en active Pending
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109916933B (en) * | 2019-01-04 | 2021-10-01 | 中国人民解放军战略支援部队信息工程大学 | X-ray computed tomography energy spectrum estimation method based on convolutional neural network |
| CN109916933A (en) * | 2019-01-04 | 2019-06-21 | 中国人民解放军战略支援部队信息工程大学 | X ray computer tomographic imaging spectra estimation method based on convolutional neural networks |
| CN109886307A (en) * | 2019-01-24 | 2019-06-14 | 西安交通大学 | A kind of image detecting method and system based on convolutional neural networks |
| CN109805963A (en) * | 2019-03-22 | 2019-05-28 | 深圳开立生物医疗科技股份有限公司 | The determination method and system of one Endometrium parting |
| CN109805963B (en) * | 2019-03-22 | 2022-07-05 | 深圳开立生物医疗科技股份有限公司 | Method and system for judging endometrium typing |
| CN110136115A (en) * | 2019-05-14 | 2019-08-16 | 重庆大学 | IVOCT image vulnerable plaque detects Approach For Neural Network Ensemble automatically |
| CN110246143A (en) * | 2019-06-14 | 2019-09-17 | 吉林大学第一医院 | Lung CT image assists detection processing device |
| CN110310281A (en) * | 2019-07-10 | 2019-10-08 | 重庆邮电大学 | Lung neoplasm detection and dividing method in a kind of Virtual Medical based on Mask-RCNN deep learning |
| CN110310281B (en) * | 2019-07-10 | 2023-03-03 | 重庆邮电大学 | Mask-RCNN deep learning-based pulmonary nodule detection and segmentation method in virtual medical treatment |
| CN110533637A (en) * | 2019-08-02 | 2019-12-03 | 杭州依图医疗技术有限公司 | A kind of method and device of test object |
| CN110533637B (en) * | 2019-08-02 | 2022-02-11 | 杭州依图医疗技术有限公司 | Method and device for detecting object |
| CN110942446A (en) * | 2019-10-17 | 2020-03-31 | 付冲 | Pulmonary nodule automatic detection method based on CT image |
| 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 |
| CN110827310A (en) * | 2019-11-01 | 2020-02-21 | 北京航空航天大学 | CT image automatic detection method and system |
| CN110838114A (en) * | 2019-11-11 | 2020-02-25 | 苏州锐一仪器科技有限公司 | Pulmonary nodule detection method, device and computer storage medium |
| CN110838114B (en) * | 2019-11-11 | 2022-07-22 | 苏州锐一仪器科技有限公司 | Pulmonary nodule detection method, device and computer storage medium |
| CN111311744A (en) * | 2020-04-02 | 2020-06-19 | 佛山市普世医学科技有限责任公司 | Candidate frame filtering, fusion and automatic updating method, system and storage medium for identifying pulmonary nodules |
| CN111798410A (en) * | 2020-06-01 | 2020-10-20 | 深圳市第二人民医院(深圳市转化医学研究院) | Cancer cell pathological grading method, device, equipment and medium based on deep learning model |
| CN112053351A (en) * | 2020-09-08 | 2020-12-08 | 哈尔滨工业大学（威海） | Method for judging benign and malignant pulmonary nodules based on neural network architecture search and attention mechanism |
| CN112529870A (en) * | 2020-12-10 | 2021-03-19 | 重庆大学 | Multi-scale CNNs (CNNs) lung nodule false positive elimination method based on combination of source domain and frequency domain |
| CN112529870B (en) * | 2020-12-10 | 2024-04-12 | 重庆大学 | Multi-scale CNNs lung nodule false positive eliminating method based on combination of source domain and frequency domain |
| CN112541909A (en) * | 2020-12-22 | 2021-03-23 | 南开大学 | Lung nodule detection method and system based on three-dimensional neural network of slice perception |
| CN112669284A (en) * | 2020-12-29 | 2021-04-16 | 天津大学 | Method for realizing pulmonary nodule detection by generating confrontation network |
| CN113361584A (en) * | 2021-06-01 | 2021-09-07 | 推想医疗科技股份有限公司 | Model training method and device, and pulmonary arterial hypertension measurement method and device |
| CN116012355A (en) * | 2023-02-07 | 2023-04-25 | 重庆大学 | Adaptive false positive lung nodule removing method based on deep learning |
| CN116012355B (en) * | 2023-02-07 | 2023-11-21 | 重庆大学 | Adaptive false positive lung nodule removing method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108986073A (en) | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame | |
| CN106504232B (en) | A kind of pulmonary nodule automatic checkout system based on 3D convolutional neural networks | |
| Chen et al. | Standard plane localization in fetal ultrasound via domain transferred deep neural networks | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| CN109003672A (en) | A kind of early stage of lung cancer detection classification integration apparatus and system based on deep learning | |
| CN108464840A (en) | A kind of breast lump automatic testing method and system | |
| Gao et al. | A deep learning based approach to classification of CT brain images | |
| CN107301640A (en) | A kind of method that target detection based on convolutional neural networks realizes small pulmonary nodules detection | |
| CN108805209A (en) | A kind of Lung neoplasm screening method based on deep learning | |
| Liu et al. | A CADe system for nodule detection in thoracic CT images based on artificial neural network | |
| CN109325942A (en) | Eye fundus image Structural Techniques based on full convolutional neural networks | |
| CN110047082A (en) | Pancreatic Neuroendocrine Tumors automatic division method and system based on deep learning | |
| CN103955698B (en) | The method of standard tangent plane is automatically positioned from ultrasonoscopy | |
| CN110838114B (en) | Pulmonary nodule detection method, device and computer storage medium | |
| Li et al. | Explainable multi-instance and multi-task learning for COVID-19 diagnosis and lesion segmentation in CT images | |
| CN107767362A (en) | A kind of early screening of lung cancer device based on deep learning | |
| Camarlinghi | Automatic detection of lung nodules in computed tomography images: training and validation of algorithms using public research databases | |
| CN105556567B (en) | Method and system for vertebral location detection | |
| Nayan et al. | A deep learning approach for brain tumor detection using magnetic resonance imaging | |
| Mei et al. | YOLO-lung: A practical detector based on imporved YOLOv4 for Pulmonary Nodule Detection | |
| Xiao et al. | A cascade and heterogeneous neural network for CT pulmonary nodule detection and its evaluation on both phantom and patient data | |
| Yang et al. | A multi-stage progressive learning strategy for COVID-19 diagnosis using chest computed tomography with imbalanced data | |
| Pei et al. | Computerized detection of lung nodules in CT images by use of multiscale filters and geometrical constraint region growing | |
| Zhang et al. | A computer aided diagnosis system in mammography using artificial neural networks | |
| CN111402231B (en) | Automatic evaluation system and method for lung CT image quality |
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: 20181211 |