CN110215232A - Ultrasonic patch analysis method in coronary artery based on algorithm of target detection - Google Patents
Ultrasonic patch analysis method in coronary artery based on algorithm of target detection Download PDFInfo
- Publication number
- CN110215232A CN110215232A CN201910358808.XA CN201910358808A CN110215232A CN 110215232 A CN110215232 A CN 110215232A CN 201910358808 A CN201910358808 A CN 201910358808A CN 110215232 A CN110215232 A CN 110215232A
- Authority
- CN
- China
- Prior art keywords
- algorithm
- detection
- patch
- loss
- target detection
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 73
- 238000004458 analytical method Methods 0.000 title claims abstract description 18
- 210000004351 coronary vessel Anatomy 0.000 title claims abstract description 15
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000002608 intravascular ultrasound Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 7
- 238000000034 method Methods 0.000 claims description 7
- 238000013526 transfer learning Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000007689 inspection Methods 0.000 claims 1
- 238000003745 diagnosis Methods 0.000 abstract description 14
- 230000015572 biosynthetic process Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 4
- 208000029078 coronary artery disease Diseases 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000002604 ultrasonography Methods 0.000 description 2
- 206010002383 Angina Pectoris Diseases 0.000 description 1
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000003340 mental effect Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004165 myocardium Anatomy 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
- 238000003325 tomography Methods 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/08—Detecting organic movements or changes, e.g. tumours, cysts, swellings
- A61B8/0891—Detecting organic movements or changes, e.g. tumours, cysts, swellings for diagnosis of blood vessels
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/12—Diagnosis using ultrasonic, sonic or infrasonic waves in body cavities or body tracts, e.g. by using catheters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/5215—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data
- A61B8/5223—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data for extracting a diagnostic or physiological parameter from medical diagnostic data
-
- 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
- G06T2207/10132—Ultrasound image
-
- 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
Abstract
The invention belongs to clinics to assist quick diagnosis applied technical field, ultrasonic patch analysis method in a kind of coronary artery based on algorithm of target detection is disclosed, data are acquired first, secondly blood vessel detection and handmarking are carried out to data, the data pre-processed input neural network is finally carried out to the training of algorithm of target detection.The present invention is analyzed compared to original object detection method, carries out the rate and accuracy that network training improves detection by using lightweight feature extraction network and limitation candidate frame formation range, the workload of doctor is reduced in practical diagnosis and treatment.The present invention is based on algorithm of target detection to judge endovascular plaque type, guarantees its real-time while guaranteeing accuracy；Surgeon stress can effectively be mitigated, improve diagnosis and treatment speed.
Description
Technical field
The invention belongs to clinical auxiliary quick diagnosis applied technical fields more particularly to a kind of based on algorithm of target detection
Ultrasonic patch analysis method in coronary artery.
Background technique
Currently, the prior art commonly used in the trade is such that intravascular ultrasound is a kind of endovascular of rising in recent years
Tomography technology plays important booster action in coronary intervention, also the blueness increasingly by intervention doctor
It looks at.Coronary heart disease is the severe cardiovascular disease that a kind of disease incidence increases year by year, and angina pectoris, cardiac muscle stalk may occur at any time for patient
Extremely even die suddenly.Intervention diagnosis and therapy plays vital effect in saving patient vitals, and intravascular ultrasound is intervention
An important tool in diagnosis and treatment, to provide important clue in patch identification and optimization operation plan.
The clinically analysis of various intravascular ultrasound images depends on manual identified, and some type of coronary heart disease risk
Height, the state of an illness is anxious, proposes great challenge to the quick and precisely diagnosis and treatment ability of doctor.In addition current domestic generally existing doctor's mistake
Degree fatigue, the excessively tight status of spirit.
In conclusion problem of the existing technology is: the problems such as existing measurement method is time-consuming and there are human errors,
Some type of coronary heart disease risk is high, and the state of an illness is anxious, proposes great challenge to the quick and precisely diagnosis and treatment ability of doctor.It is specific next
It says, crowd doctor matches wretched insufficiency at present, especially intervention doctor.Operation is often continued until the late into the night from morning, even
Midnight, there are also emergency treatment demand, this mental and physical strength for needing doctor high.However, the increasing of workload may cause error, very
To be occur malpractice.In addition, the IVUS image detection of single branch vessel may generate the image of a thousand sheets, doctor is not only wanted
Image quickly and is accurately analyzed, may also need manually to measure image to ensure the precision performed the operation, and this is undoubtedly
It needs to expend many energy again.Therefore, based on the image analysis of deep learning, help to assist doctor's rapid survey and diagnostic graph
Picture.
Solve the difficulty of above-mentioned technical problem:
The doctor for possessing data does not know about deep learning method, and the engineer for understanding deep learning method can not obtain effectively
Labeled data.
Solve the meaning of above-mentioned technical problem: the method for algorithm of target detection assistant analysis diagnosis Patch properties will mitigate
Surgeon stress plays certain effect in improving diagnosis and treatment speed.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of coronary arteries based on algorithm of target detection
Interior ultrasound patch analysis method,
The invention is realized in this way ultrasonic patch analysis side in a kind of coronary artery based on algorithm of target detection
Method includes:
Step 1: data acquisition: intravascular ultrasound dynamic video figure is obtained by intravascular ultrasound instrument, utilizes opencv pairs
Every 10 frame of video is once sampled；
Step 2: blood vessel detection: blood vessel detection is carried out to every frame image of acquisition, the blood vessel that will test zooms to identical
Ratio；
Step 3: data mark: using being manually labeled to data, by the blood vessel detected in blood vessel detection by institute
The different patch type contained is labeled using labelImg, generates the xml document of 4 dimension matrixes；
Step 4: data prediction: the data marked are formatted, and convert form for xml document
Csv file, finally combined with picture and be converted to the tfrecord format that tensorflow can be identified；
Step 5: network training: entering data into neural network, using there is the transfer learning mode of supervision to carry out
Algorithm of target detection training, the study of network parameter is carried out using adam optimization algorithm.
Further, described to enter data into neural network in step 5, using the transfer learning mode for having supervision
Carrying out algorithm of target detection training includes:
(1) feature extraction is carried out to input picture using MobileNetV2 network；
(2) conv4-3, conv-7 (FC7), conv6- is utilized when being detected i.e. detection using feature pyramid structure
These feature maps of different sizes of 2, conv7-2, conv8_2, conv9_2, generate on multiple feature maps
Priorbox carries out classification and position returns,
Further, in step (2), major networks parameter is as follows:
(1) classification and regression parameter
Patch classification is predicted by softmax, in total n+1 class, wherein n is patch classification, and 1 is background.To patch region
Position recurrence is carried out, i.e. prediction (x, y, w, h), wherein x, y is respectively the top left co-ordinate in patch region, and w, h are the patch
It is long and wide；
(2) building of loss function
Wherein
Loss has been divided into Classification Loss and has returned loss two parts, and wherein N is the quantity for being matched to the candidate frame of callout box；
And alpha parameter is used to adjust Classification Loss and returns the ratio between loss, defaults α=1；Classification in SSD algorithm of target detection
Loss is that typical softmax intersects entropy loss: and returning loss is typical smooth L1 loss；
(3) create-rule of candidate frame
Centered on the midpoint of point each on characteristic pattern, the concentric candidate frame of some column of generation (and then the coordinate of central point
Can be equivalent to from characteristic pattern position multiplied by step and map back original image position), each characteristic pattern corresponds to the size of candidate frame by following
Formula determines
First layer characteristic pattern corresponding min_size=S1, max_size=S2；The second layer min_size=S2, max_
Size=S3；Other are analogized.
Further, since the patch target of blood vessel interimage is smaller when the algorithm of target detection detects, pass through limitation
Priorbox range can achieve for the good detection effect of patch.
In conclusion advantages of the present invention and good effect are as follows: the present invention divides compared to original artificial recognition methods
Analysis carries out target detection network training by using lightweight feature extraction network and limitation candidate frame formation range and improves
The rate and accuracy of detection to be implanted into algorithm model in portable medical device, can be in practical diagnosis and treatment
Reduce the workload of doctor.The present invention is based on algorithm of target detection to judge endovascular plaque type, is guaranteeing accuracy
Guarantee its real-time simultaneously；Surgeon stress can effectively be mitigated, improve diagnosis and treatment speed.The present invention is according to the gray-scale figure of intravascular ultrasound
The characteristics of picture, realizes the detection and classification to Coronary Artery Lesions using algorithm of target detection, and then doctor is helped to examine faster
It is disconnected.The present invention is by improving SSD algorithm of target detection, using the smaller feature extraction network MobileNetV2 of magnitude to blood vessel
Interior ultrasound image carries out the speed and precision that feature extraction can accelerate training and detect；It is raw to different characteristic figure limitation candidate frame
At range, the patch in coronarography can be preferably detected compared to traditional images processing method, in practical diagnosis and treatment
In reduce the workload of doctor.The feature extraction network MobileNetV2 that the present invention uses phase in detection and training performance
Than suffering from certain raising in MobileNetV1, the time of detection training can be shortened；By adjusting the progress of candidate frame range
SSD target detection model training, can accelerate the convergence rate of model.
Detailed description of the invention
Fig. 1 is ultrasonic patch analysis side in the coronary artery provided in an embodiment of the present invention based on algorithm of target detection
Method flow chart.
Fig. 2 is ultrasonic patch analysis side in the coronary artery provided in an embodiment of the present invention based on algorithm of target detection
Method schematic diagram.
Fig. 3 is target detection schematic network structure provided in an embodiment of the present invention.
Fig. 4 is data mark schematic diagram provided in an embodiment of the present invention.
Fig. 5 is plaque type schematic diagram provided in an embodiment of the present invention.
Fig. 6 is intravascular ultrasound provided in an embodiment of the present invention-virtual histology image schematic diagram.
Fig. 7 is the model training speed provided in an embodiment of the present invention in different initialization networks.
Fig. 8 is the training performance provided in an embodiment of the present invention under different candidate frame formation ranges.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
Application principle of the invention is explained in detail with reference to the accompanying drawing.
As Figure 1-Figure 2, ultrasonic in the coronary artery provided in an embodiment of the present invention based on algorithm of target detection
Patch analysis method includes:
S101: data acquisition: intravascular ultrasound dynamic video figure is obtained by intravascular ultrasound instrument, using opencv to view
Frequently every 10 frame is once sampled；
S102: blood vessel detection: blood vessel detection is carried out to every frame image of acquisition, the blood vessel that will test zooms to mutually on year-on-year basis
Example；
S103: data mark: using being manually labeled to data, by the blood vessel detected in blood vessel detection by contained
Some different patch types are labeled using labelImg；
S104: data prediction: the data marked are formatted, and are converted to what tensorflow can be identified
Tfrecord format；
S105: network training: entering data into neural network, using there is the transfer learning mode of supervision to carry out mesh
Detection algorithm training is marked, the study of network parameter is carried out using adam optimization algorithm.
As shown in figure 3, in step S105, it is provided in an embodiment of the present invention to enter data into neural network, it uses
There is the transfer learning mode of supervision to carry out algorithm of target detection training to include:
(1) feature extraction is carried out to input picture using MobileNetV2 network；
(2) conv4-3, conv-7 (FC7), conv6- is utilized when being detected i.e. detection using feature pyramid structure
These feature maps of different sizes of 2, conv7-2, conv8_2, conv9_2, generate on multiple feature maps
Priorbox carries out classification and position returns.
In step (2), major networks parameter provided in an embodiment of the present invention includes:
1) classification and regression parameter include:
Patch classification is predicted by softmax, in total n+1 class, wherein n is patch classification, and 1 is background.To patch region
Position recurrence is carried out, i.e. prediction (x, y, w, h), wherein x, y is respectively the top left co-ordinate in patch region, and w, h are the patch
It is long and wide；
2) building of loss function
Wherein
Loss has been divided into Classification Loss and has returned loss two parts, and wherein N is the quantity for being matched to the candidate frame of callout box；
And alpha parameter is used to adjust Classification Loss and returns the ratio between loss, defaults α=1；Classification in SSD algorithm of target detection
Loss is that typical softmax intersects entropy loss: and returning loss is typical smooth L1 loss；
3) create-rule of candidate frame
Centered on the midpoint of point each on characteristic pattern, the concentric candidate frame of some column of generation (and then the coordinate of central point
Can be equivalent to from characteristic pattern position multiplied by step and map back original image position), each characteristic pattern corresponds to the size of candidate frame by following
Formula determines
First layer characteristic pattern corresponding min_size=S1, max_size=S2；The second layer min_size=S2, max_
Size=S3；Other are analogized.
Algorithm of target detection provided in an embodiment of the present invention is led to since the patch target of blood vessel interimage is smaller when detection
Crossing limitation priorbox range can achieve for the good detection effect of patch.
This law invention specific embodiment:
1, method:
Data acquisition: the first step obtains intravascular ultrasound dynamic video figure by intravascular ultrasound instrument, utilizes opencv pairs
Every 10 frame of video is once sampled.
Blood vessel detection: second step carries out blood vessel detection to every frame image of acquisition, the blood vessel that will test zooms to identical
Ratio.
Third step, data mark: using being manually labeled to data, by the blood vessel detected in blood vessel detection by institute
The different patch type contained is labeled using labelImg.
4th step, data prediction: the data marked are formatted, and are converted to what tensorflow can be identified
Tfrecord format.
5th step, network training: entering data into neural network, using there is the transfer learning mode of supervision to carry out
Algorithm of target detection training, the study of network parameter is carried out using adam optimization algorithm.
Patch is divided into composition in 4 by traditional intravascular ultrasound-virtual histology, but the application under grayscale needs to cure
The characteristics of raw concrete analysis, the present invention is the gray scale image according to intravascular ultrasound, is realized using algorithm of target detection to hat
The detection and classification of arteries and veins lesion, and then doctor is helped to diagnose faster.
Feature extraction is carried out to input picture by MobileNetV2 network, the speed of training with detection can be accelerated, with
Conv4-3, conv-7 (FC7), conv6-2, conv7- is utilized when being detected i.e. detection using feature pyramid structure afterwards
These feature maps of different sizes of 2, conv8_2, conv9_2, generate priorbox on multiple feature maps
It carries out classification and position returns, major networks parameter is as follows:
(1) classification and regression parameter
Patch classification is predicted by softmax, in total n+1 class, wherein n is patch classification, and 1 is background.To patch region
Position recurrence is carried out, i.e. prediction (x, y, w, h), wherein x, y is respectively the top left co-ordinate in patch region, and w, h are the patch
It is long and wide.
(2) building of loss function
Wherein
Loss has been divided into Classification Loss and has returned loss two parts, and wherein N is the quantity for being matched to the candidate frame of callout box；
And alpha parameter is used to adjust Classification Loss and returns the ratio between loss, defaults α=1.Classification in SSD algorithm of target detection
Loss is that typical softmax intersects entropy loss: and returning loss is typical smooth L1 loss
(3) create-rule of candidate frame
Centered on the midpoint of point each on characteristic pattern, the concentric candidate frame of some column of generation (and then the coordinate of central point
Can be equivalent to from characteristic pattern position multiplied by step and map back original image position), each characteristic pattern corresponds to the size of candidate frame by following
Formula determines
First layer characteristic pattern corresponding min_size=S1, max_size=S2；The second layer min_size=S2, max_
Size=S3；Other are analogized.
Since the patch target of blood vessel interimage is smaller, can achieve by limitation priorbox range good for patch
Detection effect.
2, specific performance is assessed
As shown in Fig. 4 to Fig. 8, the feature extraction network MobileNetV2 that the present invention uses is in detection and training performance
Certain raising is suffered from compared to MobileNetV1, can shorten the time of detection training.By adjusting candidate frame range into
Row SSD target detection model training, can accelerate the convergence rate of model.By result above it can be concluded that the present invention is led to
Cross improvement SSD algorithm of target detection, using the smaller feature extraction network MobileNetV2 of magnitude to ivus image into
The speed and precision that row feature extraction can accelerate training and detect；Candidate frame formation range is limited to different characteristic figure, is compared
The patch in coronarography can be preferably detected in traditional images processing method, reduce doctor in practical diagnosis and treatment
Workload.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (4)
1. ultrasonic patch analysis method in a kind of coronary artery based on algorithm of target detection, which is characterized in that the base
Include: in patch analysis method ultrasonic in the coronary artery of algorithm of target detection
Data acquisition: step 1 obtains intravascular ultrasound dynamic video figure by intravascular ultrasound instrument, using opencv to video
Every 10 frame is once sampled；
Blood vessel detection: step 2 carries out blood vessel detection to every frame image of acquisition, the blood vessel that will test zooms to mutually on year-on-year basis
Example；
Step 3, data mark: using being manually labeled to data, by the blood vessel detected in blood vessel detection by contained
Different patch type be labeled using labelImg；
Step 4, data prediction: the data marked are formatted, and are converted to what tensorflow can be identified
Tfrecord format；
Step 5, network training: entering data into neural network, using there is the transfer learning mode of supervision to carry out target
Detection algorithm training, the study of network parameter is carried out using adam optimization algorithm.
2. ultrasonic patch analysis method, feature in the coronary artery based on algorithm of target detection as described in claim 1
It is, it is described to enter data into neural network in step 5, using there is the transfer learning mode of supervision to carry out target inspection
Method of determining and calculating training includes:
(1) feature extraction is carried out to input picture using MobileNetV2 network；
(2) conv4-3, conv-7 (FC7), conv6-2 is utilized when being detected i.e. detection using feature pyramid structure,
These feature maps of different sizes of conv7-2, conv8_2, conv9_2, generate on multiple feature maps
Priorbox carries out classification and position returns.
3. ultrasonic patch analysis method, feature in the coronary artery based on algorithm of target detection as claimed in claim 2
It is, in step (2), the detection major networks parameter includes:
(1) classification and regression parameter
Patch classification is predicted by softmax, in total n+1 class, wherein n is patch classification, and 1 is background, is carried out to patch region
Position returns, i.e. prediction (x, y, w, h), and wherein x, y are respectively the top left co-ordinate in patch region, w, h be the patch length and
It is wide；
(2) building of loss function
Wherein
Loss has been divided into Classification Loss and has returned loss two parts, and wherein N is the quantity for being matched to the candidate frame of callout box；And α
Parameter is used to adjust Classification Loss and returns the ratio between loss, defaults α=1；Classification Loss in SSD algorithm of target detection
It is that typical softmax intersects entropy loss: and returning loss is typical smooth L1 loss；
(3) create-rule of candidate frame
Centered on the midpoint of point each on characteristic pattern, the concentric candidate frame of some column is generated, then the coordinate of central point can multiply
It with step, is equivalent to from characteristic pattern position and maps back original image position, each characteristic pattern corresponds to the size of candidate frame by following formula
It determines；
First layer characteristic pattern corresponding min_size=S1, max_size=S2；The second layer min_size=S2, max_size=
S3；Other are analogized.
4. ultrasonic patch analysis method, feature in the coronary artery based on algorithm of target detection as claimed in claim 2
It is, when the object detection method is detected, since the patch target of blood vessel interimage is smaller, by limiting priorbox
Range can achieve for the good detection effect of patch.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910358808.XA CN110215232A (en) | 2019-04-30 | 2019-04-30 | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910358808.XA CN110215232A (en) | 2019-04-30 | 2019-04-30 | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110215232A true CN110215232A (en) | 2019-09-10 |
Family
ID=67820181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910358808.XA Pending CN110215232A (en) | 2019-04-30 | 2019-04-30 | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110215232A (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107621626A (en) * | 2017-10-09 | 2018-01-23 | 中国矿业大学(北京) | Radar signal Railway Roadbed detection method based on depth convolutional neural networks |
| CN107730497A (en) * | 2017-10-27 | 2018-02-23 | 哈尔滨工业大学 | A kind of plaque within blood vessels property analysis method based on depth migration study |
| CN108182683A (en) * | 2018-02-08 | 2018-06-19 | 山东大学 | IVUS based on deep learning and transfer learning organizes mask method and system |
| CN108564097A (en) * | 2017-12-05 | 2018-09-21 | 华南理工大学 | A kind of multiscale target detection method based on depth convolutional neural networks |
| CN108985193A (en) * | 2018-06-28 | 2018-12-11 | 电子科技大学 | A kind of unmanned plane portrait alignment methods based on image detection |
| CN109146967A (en) * | 2018-07-09 | 2019-01-04 | 上海斐讯数据通信技术有限公司 | The localization method and device of target object in image |
| CN109409252A (en) * | 2018-10-09 | 2019-03-01 | 杭州电子科技大学 | A kind of traffic multi-target detection method based on modified SSD network |
| CN109584209A (en) * | 2018-10-29 | 2019-04-05 | 深圳先进技术研究院 | Vascular wall patch identifies equipment, system, method and storage medium |
| CN109685066A (en) * | 2018-12-24 | 2019-04-26 | 中国矿业大学(北京) | A kind of mine object detection and recognition method based on depth convolutional neural networks |
| CN109684920A (en) * | 2018-11-19 | 2019-04-26 | 腾讯科技（深圳）有限公司 | Localization method, image processing method, device and the storage medium of object key point |
-
2019
- 2019-04-30 CN CN201910358808.XA patent/CN110215232A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107621626A (en) * | 2017-10-09 | 2018-01-23 | 中国矿业大学(北京) | Radar signal Railway Roadbed detection method based on depth convolutional neural networks |
| CN107730497A (en) * | 2017-10-27 | 2018-02-23 | 哈尔滨工业大学 | A kind of plaque within blood vessels property analysis method based on depth migration study |
| CN108564097A (en) * | 2017-12-05 | 2018-09-21 | 华南理工大学 | A kind of multiscale target detection method based on depth convolutional neural networks |
| CN108182683A (en) * | 2018-02-08 | 2018-06-19 | 山东大学 | IVUS based on deep learning and transfer learning organizes mask method and system |
| CN108985193A (en) * | 2018-06-28 | 2018-12-11 | 电子科技大学 | A kind of unmanned plane portrait alignment methods based on image detection |
| CN109146967A (en) * | 2018-07-09 | 2019-01-04 | 上海斐讯数据通信技术有限公司 | The localization method and device of target object in image |
| CN109409252A (en) * | 2018-10-09 | 2019-03-01 | 杭州电子科技大学 | A kind of traffic multi-target detection method based on modified SSD network |
| CN109584209A (en) * | 2018-10-29 | 2019-04-05 | 深圳先进技术研究院 | Vascular wall patch identifies equipment, system, method and storage medium |
| CN109684920A (en) * | 2018-11-19 | 2019-04-26 | 腾讯科技（深圳）有限公司 | Localization method, image processing method, device and the storage medium of object key point |
| CN109685066A (en) * | 2018-12-24 | 2019-04-26 | 中国矿业大学(北京) | A kind of mine object detection and recognition method based on depth convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 赵媛 等: "一种基于深度学习的颈动脉斑块超声图像识别方法", 《中国医疗器械信息》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103298412B (en) | Pulmonary blood pressure is measured using through chest lung doppler ultrasound | |
| Aggarwal et al. | In-vivo heterogeneous functional and residual strains in human aortic valve leaflets | |
| CN109512464A (en) | A kind of disorder in screening and diagnostic system | |
| CN108665449B (en) | Deep learning model and device for predicting blood flow characteristics on blood flow vector path | |
| KR20200006447A (en) | Method and program for stroke volume acquisition using arterial pressure waveform | |
| Pedrizzetti et al. | Functional strain-line pattern in the human left ventricle | |
| Zhou et al. | Automatic myotendinous junction tracking in ultrasound images with phase-based segmentation | |
| Sommer et al. | Patient-specific 3D-printed coronary models based on coronary computed tomography angiography volumes to investigate flow conditions in coronary artery disease | |
| CN110353639A (en) | A kind of blood supply area quantitative approach and system based on blood vessel enhancing radiography | |
| CN116778021B (en) | Medical image generation method, device, electronic equipment and storage medium | |
| CN110310364B (en) | Heart 3D printing system and printing method thereof | |
| CN110215232A (en) | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection | |
| TWI590802B (en) | System and method for arrhythmia judgment data base and ecg arrhythmia judgment | |
| Sabia et al. | Evaluation of aortic stiffness by a new simplified 2D speckle tracking analysis | |
| CN113592802B (en) | Mitral valve annular displacement automatic detection system based on ultrasonic image | |
| CN111466952B (en) | Real-time conversion method and system for ultrasonic endoscope and CT three-dimensional image | |
| JP5885234B2 (en) | Image analysis method in disease determination apparatus and ultrasonic image forming apparatus | |
| Shaaf et al. | A Convolutional Neural Network Model to Segment Myocardial Infarction from MRI Images. | |
| CN114176532A (en) | Clinical verification method for determining cfPWV parameters and application system thereof | |
| CN112914610A (en) | Contrast-enhanced echocardiography wall thickness automatic analysis system and method based on deep learning | |
| Hagenah et al. | Generating healthy aortic root geometries from ultrasound images of the individual pathological morphology using deep convolutional autoencoders | |
| CN113017667A (en) | Method, device and equipment for quantifying vascular stenosis and readable storage medium | |
| CN110837751A (en) | Human motion capture and gait analysis method based on RGBD depth camera | |
| Liu et al. | Central Aortic Blood Pressure Waveform Estimation Based on Temporal Convolutional Network | |
| CN111281372B (en) | Method, device and system for determining heart failure change index |
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: 20190910 |
|
| RJ01 | Rejection of invention patent application after publication |