CN110264465A - A kind of dissection of aorta dynamic testing method based on morphology and deep learning - Google Patents
A kind of dissection of aorta dynamic testing method based on morphology and deep learning Download PDFInfo
- Publication number
- CN110264465A CN110264465A CN201910564318.5A CN201910564318A CN110264465A CN 110264465 A CN110264465 A CN 110264465A CN 201910564318 A CN201910564318 A CN 201910564318A CN 110264465 A CN110264465 A CN 110264465A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- image
- aorta
- dissection
- cta
- 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
- G06T2207/10—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
- G06T2207/20—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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
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)
- Image Processing (AREA)
Abstract
Dissection of aorta (Aortic dissection (AD)) is a kind of cardiovascular disease of danger, has high clinically dead rate, and incidence is increased sharply in recent years；The invention discloses a kind of dissection of aorta dynamic testing method based on morphology and deep learning, step 1: establishing CTA sample set；Step 2: deep learning being carried out based on the CTA image in CTA sample set and implements disease detection；2.1 carry out deep learning, the convolutional neural networks after being learnt based on area-of-interest；2.2 are directed to new CTA image, based on the convolutional neural networks after study and area-of-interest are combined to carry out disease detection.The dissection of aorta dynamic testing method based on morphology and deep learning is easy to implement, and detection efficiency is high.Experiment shows to be far superior to conventional method using deep learning method, and the method based on DenseNet121 is more excellent.
Description
Technical field
The present invention relates to a kind of dissection of aorta dynamic testing method based on morphology and deep learning.
Background technique
Dissection of aorta (Aortic dissection (AD)) is a kind of cardiovascular disease of danger, has high face
The bed death rate, and incidence is increased sharply in recent years；This disease is originating from aorta wall (aortic wall) inner membrance and middle layer
It tears (intimal tears)；Aorta intracavity blood is directed through in lesion under the driving of pulse pressure through inner membrance tearing port
Layer, causes middle layer to separate；In recent years, the pathogenetic case of this disease is growing day by day.
Slice CT Angiography art is a kind of angiography of non-dispersive, by inject in the blood vessel contrast agent to
Show the blood vessel of different physical feelings, which can not only provide the shape information of lumen variation, moreover it is possible to show the tube wall of blood vessel
Lesion；With safety, repeat, the advantages such as accuracy height.
However, each patient needs to obtain 300-500 width CTA image；
Analysis image needs to spend the time of many hours of expert, since the experience of each expert is different, diagnosis knot
Fruit otherness is larger, it is easy to lead to wrong diagnosis.
In recent years, medical image auxiliary diagnosis is increasingly applied, and still, traditional machine learning method can not be competent at sea
The image procossing of amount.However, human brain can be simulated by deep learning, and high-efficient.Based on convolutional neural networks
(convolutional neural networks, CNN) application is increasingly extensive.
But existing method accomplishes to be widely popularized not yet since accuracy rate is not high.
Therefore, it is necessary to design a kind of new dissection of aorta dynamic testing method based on morphology and deep learning.
Summary of the invention
It is dynamic that technical problem to be solved by the invention is to provide a kind of dissection of aorta based on morphology and deep learning
State detection method, the dissection of aorta dynamic testing method based on morphology and deep learning is easy to implement, and method is unique, quasi-
True property is high.
The technical solution of invention is as follows:
A kind of dissection of aorta dynamic testing method based on morphology and deep learning, which is characterized in that
Step 1: establishing CTA sample set；
Step 2: deep learning being carried out based on the CTA image in CTA sample set and implements disease detection；
2.1 carry out deep learning, the convolutional neural networks after being learnt based on area-of-interest；
2.2 are directed to new CTA image, based on the convolutional neural networks after study and area-of-interest are combined to carry out disease inspection
It surveys.
In step 1, the CTA sample set including 88 samples is established, including 45 healthy samples and 43
Disease sample；The sample has 4840 sectioning images；Above-mentioned samples sources are in refined attached second hospital in Hunan, using multiple experiences
Cardiovascular medicine expert abundant classifies to sample set, that is, is divided into healthy sample and disease sample.
The step of establishing area-of-interest (the region ofinterests, ROI) of CTA image are as follows:
Step a: image binaryzation；
Image is carried out by binaryzation using adaptive OSTU method；
Original image is divided into 8*8 sub-block, is separated the image in each block using OSTU method；Final choosing
Maximum sub-block segmentation threshold is selected as final segmentation threshold, obtained binary map；
Step b: opening operation processing；
Opening operation processing is carried out to binary map, removes the corresponding part in abdominal cavity in image；
Step c:ROI is positioned and is formed area-of-interest；
With i block, the center point coordinate of i-th of block is (x (i), y (i))；
Note x (i) and .y (i) are the abscissa and ordinate of i-th of central point respectively, then the central point of ROI are as follows:
X_c and y_c is the abscissa and ordinate of ROI region；
Central point based on ROI is determined position and the size of area-of-interest by preset rectangular dimension；
ROI is directly mapped in original image, the image of area-of-interest is obtained.
Deep learning is carried out using DcnseNet121 method.
The utility model has the advantages that
Dissection of aorta dynamic testing method based on morphology and deep learning of the invention, by establishing a sample
This collection, and disease detection is carried out based on sample set, using deep learning network (InceptionV3, ResNet50, DenseNet)
It is tested, and uses Recall, F1-score, MCC and other indexs carry out assessment experiment and show deep learning method much
Better than conventional method, and the method based on DenseNet121 is more excellent.
Main contributions of the invention are:
Firstly, the CTA sample set including 88 samples is established, including 45 healthy samples and 43 diseases
Sample；These samples sources divide it using label in hospital clinical data, 2 veteran cardiovascular medicine experts
Class；
Second, disease detection is carried out based on CTA image, deep learning method is applied into the area-of-interest in sample
(region of interests, ROI)；
Finally, analyzing 4840 sectioning images of 88 all samples, using conventional method and deep learning side
Method comparison, detailed comparisons' analysis use distinct methods in the difference of sensibility and accuracy；Deep learning method includes
DenseNet, Resnet and InceptionV3.
Detailed description of the invention
Fig. 1 is CTA dissection of aorta schematic diagram (dissection surface of the arch of aorta)；
Fig. 2 is CTA dissection of aorta schematic diagram (dissection surface of aorta ascendens and descending aorta)；
Fig. 3 is the broad flow diagram of the method for the present invention；
Fig. 4 is CTA image original image；
Fig. 5 is by the image after common OSTU method binaryzation；
Fig. 6 is the schematic diagram divided the image into after 8X8 sub-block；
Fig. 7 is based on the binary map after piecemeal binaryzation；
Fig. 8 is opening operation effect diagram；
Fig. 9 is that ROI positions schematic diagram；
Figure 10 is the structural schematic diagram of DenseNet121 convolutional network；
Figure 11 is each method comparison schematic diagram.
Specific embodiment
The present invention is described in further details below with reference to the drawings and specific embodiments:
Embodiment 1: the basic principle of dissection of aorta disease:
Dissection of aorta disease refers to aorta intracavity blood under the driving of pulse pressure, is directed through disease through inner membrance tearing port
Become middle layer, middle layer is caused to separate；And it moves towards to extend along artery, forms true and false chamber.
The figure of the CTA with dissection of aorta is shown in Fig. 1-2；With the change of scan position, the shape meeting of aorta
It changes.Fig. 1, in 2, superior vena cava refers to that superior vena cava, trachea are tracheae, Dssection of
Aortic arch is arch of aorta interlayer；Dissection of ascending aorta is ascending artery interlayer；left
Coronory artery is arteria coroaria sinistra；Pulmonary artery is pulmonary artery；dissection of descending
Aorta is descending artery interlayer.
Fig. 1 shows the sectional view of arch of aorta position (aortic arch), with the enclosed part of yellow line in figure
It is exactly the arch of aorta with dissection of aorta.Fig. 2 shows the sectional view of aorta ascendens and descending aorta position, in figure
It is the part to be fenced up according to lines that 2 regions of mark, which respectively correspond aorta ascendens and the specific Diseases diagnosis of descending aorta,
Come what is implemented.
Establish data set
CTA image is collected from hospital clinical and is used for training and detection of the invention, specifically, these CTA images are to be directed to
Aorta ascendens, descending aorta and artery arch position.Image is divided into 4 classes, referring to table 1；In total from 88 CTA cases
4840 sectioning images, in 88 cases, including 43 dissection of aorta patients and 45 normal cases.It is also shown in table 1
Specific quantity and ratio.
Table 1: the structure of data set
OSTU identification, this method are existing maturation method.
OSTU method [Otsu (1979)] is the image identification method based on threshold value.This method is obtained optimal by grayscale image
Threshold value.
Assuming that M0And M1Respectively foreground and background, probability are as follows:
piIt is the probability that pixel value is i, ξ (k) is the probability that pixel value is 1-k；
L is the grey level of whole image, value 256；Then average gray are as follows:
μ (k) is the average gray that pixel value is k；The class variances can be given by:
μ0, ξ0The respectively average gray and probability of prospect；
So thatIt is maximum, so that it may to obtain optimal threshold.μ1ξ1, it is the average gray and probability for background respectively.
The quasi- method taken
The method of use includes 2 parts: ROI (area-of-interest) is extracted and the deep learning based on detection, referring to fig. 2
ROI (area-of-interest) is extracted for CTA image, and intermediate arterial portions are most important and dissection of aorta
Core, in order to enable interference of other regions to detection, needs to extract ROI.
Image binaryzation
One complete CTA image can be divided into 2 parts: the information area and abdominal cavity region；Information area includes the private of patient
People's information and facility information, the region have the black background of large area, and will not influence disease detection；And abdominal cavity region is shown
Intermediate aorta, therefore, binaryzation are used to distinguish the foreground and background of image；However, abdominal cavity (abdominal
Cavity the gray value of gray value) and the gray value of arterial portions are similar, therefore, it is difficult to which the two is distinguished；In fact,
In OSTU method, abdominal cavity will maximum probability be divided into prospect, referring to Fig. 5.
Adaptive OSTU method is suggested, and in CTA image processing process, is aimed at arteriosomes are complete with abdominal cavity
It is fully separating；Such as Fig. 6, original image is divided into 8*8 sub-block, is separated the image in each block using OSTU method；
The maximum sub-block separation threshold value of final choice is as final separation threshold value, obtained binary map such as Fig. 7；
The main program of piecemeal OSTU algorithm:
Opening operation
After binaryzation, the boundary of some noises and abdominal cavity is still had in image；Therefore, it is necessary to use mathematical morphology
Opening operation remove these influence of noises.
In mathematical morphology, be used to describe essential characteristic or basic structure there are sequence of operations, most basic corrosion and
Expansive working.
Corrosion is used to cut down target area, expression formula are as follows:
A is original image, and B is the construction operator for corrosion image A.
It expands for expanding the boundary of target, fills some duck eyes, expression formula are as follows:
A is original image, and B is the construction operator for expanding image A.
Opening operation, which refers to, first corrodes reflation, expression formula are as follows:
A is original image, and B is the construction operator for converting image A.
By opening operation, more smaller than construction operator isolated dot and burr are removed.
Also, the connection between each image block is interrupted, and there is no change for the shape of general image.
It such as Fig. 8, is handled by opening operation, the corresponding part in abdominal cavity is all removed from image.
ROI positioning
When carrying out ROI positioning, the block for being less than threshold value is removed as noise, remaining relevant portion is then interested
Region.
The central point of interesting part is calculated, central point may more than one.
Note x (i) and y (i) are the abscissa and ordinate of i-th of central point respectively, then the central point of ROI are as follows:
Having 1≤i≤L, L is the quantity of associated part.X_c and y_c is the abscissa and ordinate of ROI region, will
ROI is directly mapped in original image, referring to Fig. 9.
During this, still existed in the picture by the part of non-master artery, these problems will be in next step, that is, depth
It is solved in study.
Detection based on convolutional neural networks (DenseNet)
Since 2014, nerual network technique was continued to develop, and structure is bigger, and level is more, and model originally is using tool
There is various sizes of multiple convolution, subsequent model V2 and V3 (InceptionV2 and IncepionV3), which have, more preferably to be showed；Mould
In type V2, the convolution kernel for replacing one 5 × 5 with 23 × 3 convolution kernels (convolution kernels) enhances CNN's
Learning ability.
Meanwhile batch processing normalization (batch normalization (BN)) method can further speed up training process.?
In model V3, the thought of small convolution factor is further applied, the phenomenon that for reducing overfitting.
In addition, the series with network increases, training difficulty is consequently increased；Residual error learning network (residual
Learning (ResNet)) it is applied, by the way that input information is directly converted to output information, the integrality of information is protected
Shield.Entire study need to only consider to output and input between difference, so as to simplify learning objective and reduce learning difficulty.
Studies have shown that the overall performance of network can be significantly improved by reducing certain layers；Finally, the present invention uses intensive convolutional Neural net
Network (Densely connected convolutional network (DenseNet)), before each layer has with other layers
Feedforward gain connection, that is, increase L (L+2)/2 direct forward connection, as shown in Figure 10.In Figure 10, convolution is volume
Lamination, pooling refer to pooling layers, and prediction aortic dissection is prediction dissection of aorta.
The transmitting of this structure energy Enhanced feature, and feature is reused, in example, used using DenseNet121
In based on test set to complete final dissection of aorta detection.
Experimental result and analysis
Test the device configuration used: Intel (R) Core (TM) i7-6500X CPU@2.50GHz and 16.00GB RAM.
Keras framework is used when deep learning.
Data set includes 4840 sectioning images of 88 cases, and see Table 1 for details；For training, 20% is used for 60% image
Confirmation, 20% for testing.
These three methods of InceptionV3, ResNet50 and DenseNet121 are for testing and comparing.
The fine tuning parameter of three kinds of methods such as table 2.
Table 2: fine tuning parameter
F1 score (F1-score) and Ma Xiusi related coefficient Matthews correlation coefficient
(MCC) it is used to evaluate the method being related to:
TP is used to indicate the quantity of genuine positive value (true positive)；FP indicates false positive value (false
Positive quantity)；FN indicates genuine negative value (false negative), and TN indicates genuine negative value (true negative).
Fig. 7 illustrates the comparison of this 3 kinds of methods.As can be seen from Figure 11, DenseNet121 method has optimal recall rate (Recall),
F1 score and MCC；Susceptibility (sensitivity) is 82.18%, far super other methods.
In order to further study the difference of each method, PRE is defined are as follows:
The result of every one kind is referring to table 3. it can thus be seen that ResNet50 has optimal inspection for 0 grade of CAT image
Survey effect.However for 1-3 grades, DenseNet121 has better effect.
Table 3: various types of other performance is compared
Since data set includes multiple classifications, performance is compared using the average value of both macro and micro, is had:
Weighted average (Weighted average) is further used for assessing every a kind of example；See Table 4 for details
Table 4: relative performance compares
As shown in Table 4, DenseNet121 method has better comprehensive performance.
Conclusion: by experiment it is found that deep learning method has preferably performance than other conventional methods；Also, in depth
It spends in learning method, DenseNet121 is more preferably than other networks such as ResNet50 and InceptionV3.In addition, in the present invention
Binarization method and deep learning method are existing mature technology.
Claims (4)
1. a kind of dissection of aorta dynamic testing method based on morphology and deep learning, which is characterized in that
Step 1: establishing CTA sample set；
Step 2: deep learning being carried out based on the CTA image in CTA sample set and implements disease detection；
2.1 carry out deep learning, the convolutional neural networks after being learnt based on area-of-interest；
2.2 are directed to new CTA image, based on the convolutional neural networks after study and area-of-interest are combined to carry out disease detection.
2. the dissection of aorta dynamic testing method according to claim 1 based on morphology and deep learning, feature
It is, in step 1, the CTA sample set including 88 samples is established, including 45 healthy samples and 43 diseases
Sample；The sample has 4840 sectioning images；Above-mentioned samples sources are in hospital clinical data, using multiple veteran hearts
Vascular medicine expert classifies to sample set, that is, is divided into healthy sample and disease sample.
3. the dissection of aorta dynamic testing method according to claim 1 based on morphology and deep learning, feature
The step of being, establishing area-of-interest (region of interests, ROI) of CTA image are as follows:
Step a: image binaryzation；
Image is carried out by binaryzation using adaptive OSTU method；
Original image is divided into 8*8 sub-block, is separated the image in each block using OSTU method；Final choice is most
Big sub-block segmentation threshold is as final segmentation threshold, obtained binary map；
Step b: opening operation processing；
Opening operation processing is carried out to binary map, removes the corresponding part in abdominal cavity in image；
Step c:ROI is positioned and is extracted area-of-interest；
With i block, the center point coordinate of i-th of block is (x (i), y (i))；
Note x (i) and y (i) are the abscissa and ordinate of i-th of central point respectively, then the central point of ROI are as follows:
X_c and y_c is the abscissa and ordinate of ROI region；
Central point based on ROI is determined position and the size of area-of-interest by preset rectangular dimension；
ROI is directly mapped in original image, the image of area-of-interest is obtained.
4. the dissection of aorta dynamic detection side according to claim 1-3 based on morphology and deep learning
Method, which is characterized in that deep learning is carried out using DenseNet121 method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910564318.5A CN110264465A (en) | 2019-06-25 | 2019-06-25 | A kind of dissection of aorta dynamic testing method based on morphology and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910564318.5A CN110264465A (en) | 2019-06-25 | 2019-06-25 | A kind of dissection of aorta dynamic testing method based on morphology and deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110264465A true CN110264465A (en) | 2019-09-20 |
Family
ID=67922129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910564318.5A Pending CN110264465A (en) | 2019-06-25 | 2019-06-25 | A kind of dissection of aorta dynamic testing method based on morphology and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110264465A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112561871A (en) * | 2020-12-08 | 2021-03-26 | 中国医学科学院北京协和医院 | Aortic dissection method and device based on flat scanning CT image |
| WO2022000977A1 (en) * | 2020-06-29 | 2022-01-06 | 苏州润迈德医疗科技有限公司 | Deep learning-based aortic image acquisition system |
| KR20220039307A (en) * | 2020-09-22 | 2022-03-29 | 한림대학교 산학협력단 | Control method, device and program of system for determining aortic dissection based on non-contrast computed tomography images using artificial intelligence |
| CN114782340A (en) * | 2022-04-09 | 2022-07-22 | 中南大学湘雅二医院 | Aortic dissection detection method, system, electronic device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203432A (en) * | 2016-07-14 | 2016-12-07 | 杭州健培科技有限公司 | A kind of localization method of area-of-interest based on convolutional Neural net significance collection of illustrative plates |
| CN107610773A (en) * | 2017-09-12 | 2018-01-19 | 北京即刻叁维数据科技股份有限公司 | A kind of vascular dissection aided diagnosis method based on sustainer medical image |
| CN109086836A (en) * | 2018-09-03 | 2018-12-25 | 淮阴工学院 | A kind of automatic screening device of cancer of the esophagus pathological image and its discriminating method based on convolutional neural networks |
| CN109448854A (en) * | 2018-09-14 | 2019-03-08 | 慧影医疗科技（北京）有限公司 | A kind of construction method of pulmonary tuberculosis detection model and application |
| CN109919961A (en) * | 2019-02-22 | 2019-06-21 | 北京深睿博联科技有限责任公司 | A kind of processing method and processing device for aneurysm region in encephalic CTA image |
-
2019
- 2019-06-25 CN CN201910564318.5A patent/CN110264465A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203432A (en) * | 2016-07-14 | 2016-12-07 | 杭州健培科技有限公司 | A kind of localization method of area-of-interest based on convolutional Neural net significance collection of illustrative plates |
| CN107610773A (en) * | 2017-09-12 | 2018-01-19 | 北京即刻叁维数据科技股份有限公司 | A kind of vascular dissection aided diagnosis method based on sustainer medical image |
| CN109086836A (en) * | 2018-09-03 | 2018-12-25 | 淮阴工学院 | A kind of automatic screening device of cancer of the esophagus pathological image and its discriminating method based on convolutional neural networks |
| CN109448854A (en) * | 2018-09-14 | 2019-03-08 | 慧影医疗科技（北京）有限公司 | A kind of construction method of pulmonary tuberculosis detection model and application |
| CN109919961A (en) * | 2019-02-22 | 2019-06-21 | 北京深睿博联科技有限责任公司 | A kind of processing method and processing device for aneurysm region in encephalic CTA image |
Non-Patent Citations (1)
| Title |
|---|
| XIN-YU JIN 等: "Pulmonary nodule detection based on CT images using Convolution neural network", 《2016 9TH INTERNATIONAL SYMPOSIUM ON COMPUTATIONAL INTELLIGENCE AND DESIGN (ISCID)》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022000977A1 (en) * | 2020-06-29 | 2022-01-06 | 苏州润迈德医疗科技有限公司 | Deep learning-based aortic image acquisition system |
| WO2022000976A1 (en) * | 2020-06-29 | 2022-01-06 | 苏州润迈德医疗科技有限公司 | Aorta obtaining method based on deep learning, and storage medium |
| KR20220039307A (en) * | 2020-09-22 | 2022-03-29 | 한림대학교 산학협력단 | Control method, device and program of system for determining aortic dissection based on non-contrast computed tomography images using artificial intelligence |
| KR102423048B1 (en) * | 2020-09-22 | 2022-07-19 | 한림대학교 산학협력단 | Control method, device and program of system for determining aortic dissection based on non-contrast computed tomography images using artificial intelligence |
| CN112561871A (en) * | 2020-12-08 | 2021-03-26 | 中国医学科学院北京协和医院 | Aortic dissection method and device based on flat scanning CT image |
| CN114782340A (en) * | 2022-04-09 | 2022-07-22 | 中南大学湘雅二医院 | Aortic dissection detection method, system, electronic device and storage medium |
| CN114782340B (en) * | 2022-04-09 | 2024-03-26 | 中南大学湘雅二医院 | Aortic dissection detection method, aortic dissection detection system, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kumar et al. | An automated early diabetic retinopathy detection through improved blood vessel and optic disc segmentation | |
| CN110264465A (en) | A kind of dissection of aorta dynamic testing method based on morphology and deep learning | |
| CN109035255B (en) | Method for segmenting aorta with interlayer in CT image based on convolutional neural network | |
| CN107563983A (en) | Image processing method and medical imaging devices | |
| US20220284584A1 (en) | Computerised tomography image processing | |
| CN109087302A (en) | A kind of eye fundus image blood vessel segmentation method and apparatus | |
| CN101667289B (en) | Retinal image segmentation method based on NSCT feature extraction and supervised classification | |
| CN106570871B (en) | Fuzzy C-mean algorithm carotid ultrasound image Internal-media thickness measurement method and system | |
| CN110648344B (en) | Diabetes retinopathy classification device based on local focus characteristics | |
| CN108511055A (en) | Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule | |
| CN107204001B (en) | Automatic midium segmentation method in carotid artery ultrasonic image | |
| Zhang et al. | Hierarchical detection of red lesions in retinal images by multiscale correlation filtering | |
| CN113408603A (en) | Coronary artery stenosis degree identification method based on multi-classifier fusion | |
| Yang et al. | Label refinement with an iterative generative adversarial network for boosting retinal vessel segmentation | |
| CN115661185A (en) | Fundus image blood vessel segmentation method and system | |
| Purwanithami et al. | Hemorrhage diabetic retinopathy detection based on fundus image using neural network and FCM segmentation | |
| Indu et al. | Diagnosis of lung cancer nodules in ct scan images using fuzzy neural network | |
| Wong-od et al. | Automated segmentation of media-adventitia and lumen from intravascular ultrasound images using non-parametric thresholding | |
| Eldin et al. | Automatic Detection of Diabetic Retinopathy using Neural Networks and Support Vector Machine | |
| Elzennary et al. | Early Deep Detection for Diabetic Retinopathy | |
| Duque et al. | Breast lesions detection in digital mammography: An automated pre-diagnosis | |
| Lima et al. | Mask overlaying: a deep learning approach for individual optic cup segmentation from fundus image | |
| Aboulwafa et al. | A Transfer Learning Approach for the Automatic Detection of Fatty Liver from Ultrasound Images | |
| Mounika et al. | A Deep Hybrid Neural Network Model to Detect Diabetic Retinopathy from Eye Fundus Images | |
| Akram et al. | Diabetic Retinopathy Detection Techniques: A Review |
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 |
Application publication date: 20190920 |
|
| RJ01 | Rejection of invention patent application after publication |