CN110136810B - Analysis method of myocardial ischemia coronary blood flow reserve - Google Patents
Analysis method of myocardial ischemia coronary blood flow reserve Download PDFInfo
- Publication number
- CN110136810B CN110136810B CN201910507174.XA CN201910507174A CN110136810B CN 110136810 B CN110136810 B CN 110136810B CN 201910507174 A CN201910507174 A CN 201910507174A CN 110136810 B CN110136810 B CN 110136810B
- Authority
- CN
- China
- Prior art keywords
- myocardial
- image
- blood flow
- module
- densenet
- 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.)
- Active
Links
- 230000017531 blood circulation Effects 0.000 title claims abstract description 42
- 208000031225 myocardial ischemia Diseases 0.000 title claims abstract description 11
- 238000004458 analytical method Methods 0.000 title claims abstract description 7
- 230000002107 myocardial effect Effects 0.000 claims abstract description 34
- 238000000034 method Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 238000005070 sampling Methods 0.000 claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 230000010016 myocardial function Effects 0.000 claims abstract description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000000284 resting effect Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 5
- 241000288105 Grus Species 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 3
- 238000009792 diffusion process Methods 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 2
- 230000000302 ischemic effect Effects 0.000 claims 3
- 238000012952 Resampling Methods 0.000 claims 1
- 230000006870 function Effects 0.000 abstract description 9
- 230000004087 circulation Effects 0.000 abstract description 3
- 239000000284 extract Substances 0.000 abstract description 3
- 230000002950 deficient Effects 0.000 abstract 1
- 238000000605 extraction Methods 0.000 abstract 1
- 238000011084 recovery Methods 0.000 abstract 1
- 238000002600 positron emission tomography Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 3
- 239000003814 drug Substances 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000034994 death Effects 0.000 description 2
- 231100000517 death Toxicity 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 208000010125 myocardial infarction Diseases 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 206010058558 Hypoperfusion Diseases 0.000 description 1
- 206010042434 Sudden death Diseases 0.000 description 1
- 206010000891 acute myocardial infarction Diseases 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000004351 coronary vessel Anatomy 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002526 effect on cardiovascular system Effects 0.000 description 1
- 230000000004 hemodynamic effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 230000007102 metabolic function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000010412 perfusion Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
- 239000002759 woven fabric Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
-
- G06T5/70—
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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
- G06T2207/30104—Vascular flow; Blood flow; Perfusion
Abstract
The invention discloses an analysis method of myocardial ischemia coronary blood flow reserve, the algorithm model firstly uses data enhancement and image preprocessing, the preprocessed myocardial dynamic PET three-dimensional (3D) image uses the DenseNet architecture of multiple Convolution Neural Network (CNN) to respectively extract image space characteristics and carry out down-sampling, then another kind of algorithm model is used: the GRU (Gate recovery Unit) performs dynamic time sequence feature extraction on the 3D image features extracted in the previous step, and then forms an analyzable 3D quantitative parameter map through the upsampled DenseNet: myocardial blood flow rate (MBF). Then, according to images of PET in two states of myocardial function rest and stress, the MBF obtained by the method further calculates Coronary Flow Reserve (CFR), thereby helping a doctor to evaluate whether the myocardial and coronary circulation functions are defective or not.
Description
Technical Field
The invention relates to the technical field of medicine, in particular to an analysis method of myocardial ischemia coronary blood flow reserve.
Background
Cardiovascular deaths account for about 44% of all deaths, with acute myocardial infarction (myocardial infarction) being of most concern, occurring suddenly and with high sudden death rates. Myocardial blood flow rate (MBF) is an extremely important indicator of circulatory metabolic function of the major blood vessels of the heart. The coronary flow reserve fraction (CFR) mainly refers to the maximum capacity of the increase in coronary blood flow as the metabolic demand of the body increases, i.e. the ratio of the blood flow rate at maximum filling of the coronary blood flow (MBF in loaded state) to the blood flow rate in the basal state (MBF in resting state). CFR reflects the maximal blood flow reserve capacity of coronary circulation, is one of important indexes of coronary circulation function, and has a normal value of 3-5, and CFR (woven fabric) 2 indicates obvious myocardial hypoperfusion. The dynamic enhanced image of Positron Emission Tomography (PET) or CT/MR is mainly used clinically to realize the determination of the blood flow velocity of myocardial perfusion, but the method is limited by the complexity and time consumption of image reconstruction and hemodynamic post-processing technology, and has strong subjectivity and low consistency.
Disclosure of Invention
The invention aims to provide a method for analyzing myocardial ischemia coronary blood flow reserve, and aims to solve the problems.
The invention provides the following technical scheme: an analysis method of myocardial ischemia coronary blood flow reserve is based on multiple deep convolutional neural networks DenseNet and GRU, and comprises the following steps:
s1, preprocessing an input image to obtain a myocardial dynamic PET three-dimensional image;
s2, respectively extracting image space characteristics and performing down-sampling on the image preprocessed in the S1 by using a DenseNet architecture of a multiple convolutional neural network;
step S3, extracting dynamic time sequence characteristics of the image space characteristics extracted in the step S2 by using GRUs;
s4, forming an analyzed three-dimensional quantitative parameter map for the image space characteristics processed in the step S3 through an up-sampling DenseNet architecture to obtain a myocardial blood flow velocity MBF;
and S5, obtaining the coronary blood flow reserve fraction CFR according to the images of the PET in the resting state and the loading state of the myocardial function and the myocardial blood flow velocity MBF obtained in the step S4.
The method further comprises the following steps:
a preprocessing module for implementing the step S1;
a DenseNet convolutional neural network module to implement the step S2;
a GRU module to implement said step S3;
an upsampling module for implementing the step S4;
and communication connection is established among the preprocessing module, the DenseNet convolutional neural network module, the GRU module and the up-sampling module and is used for data transmission.
Further, the preprocessing module comprises the following steps:
step a, the resampled three-dimensional volume of a training data set is 128 multiplied by n pixels;
b, using Gaussian filtering smoothing and anisotropic diffusion filtering pretreatment;
step c, further enhancing data of the filtered image, randomly turning/rotating the y axis and the z axis at each angle, and finely adjusting the contrast; randomly moving the time sequence t by-5 to +5 time frame numbers, wherein the total frame number is unchanged;
wherein n = z-axis image layer number, t = dynamic acquisition time sequence total frame number, which can be set according to different PET acquisition sequences.
Further, the DenseNet convolutional neural network module comprises the following steps:
d, carrying out spot processing on each time frame image according to a time sequence on the preprocessed image to form a plurality of 32 multiplied by n three-dimensional spot samples, and arranging a first frame to a last frame according to a default frame number to be used as a synchronous information channel input;
and e, using a multilayer DenseNet-Block convolutional neural network, and processing all three-dimensional patches on each frame of image layer by using a convolution algorithm kernel with the filtering size of 3 multiplied by 3.
Further, the GRU module recursively updates the existing time sequence for the extracted image spatial features, so that each layer is in a form of cyclic update and forward output as an input of a next unit.
Further, the up-sampling module up-samples the three-dimensional time-series characteristics of the GRU using a multi-layer deconvolution calculation.
Further, in step S5, the gray scale map of the output myocardial blood flow velocity MBF is converted into a color code parameter map, and then the loading myocardial blood flow velocity: and obtaining the coronary blood flow reserve fraction by the resting myocardial blood flow rate.
The invention has the beneficial effects that:
aiming at a 4D time sequence image of dynamic PET myocardial imaging, a multi-convolution neural network is utilized to quantify and visualize a myocardial blood flow velocity MBF parameter map, and an artificial intelligence deep learning image processing technology is adopted to carry out automatic calculation and parameter map visualization of MBF, so that myocardial CFR is further calculated, and assessment of myocardial ischemia and myocardial infarction risks is accurately and efficiently improved.
Drawings
FIG. 1 is a schematic diagram illustrating a development process of an algorithm model according to an embodiment of the present invention;
FIG. 2 is a diagram of an algorithm module design according to an embodiment of the present invention;
fig. 3 is a diagram illustrating a timing relationship of GRU processing according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-3, an analysis method of myocardial ischemia coronary blood flow reserve based on multiple deep convolutional neural networks DenseNet and GRU includes the following steps:
s1, preprocessing an input 4D image, wherein the preprocessing process comprises image preprocessing and data enhancement to obtain a myocardial dynamic PET three-dimensional image;
s2, respectively extracting image space characteristics and performing down-sampling on the image preprocessed in the S1 by using a DenseNet architecture of a multiple convolutional neural network;
step S3, extracting dynamic time sequence characteristics of the image space characteristics extracted in the step S2 by using GRUs;
s4, forming an analyzed three-dimensional quantitative parameter map for the image space characteristics processed in the step S3 through an up-sampling DenseNet architecture to obtain a myocardial blood flow velocity MBF;
and S5, obtaining the coronary blood flow reserve fraction CFR according to the images of the PET in the resting state and the loading state of the myocardial function and the myocardial blood flow velocity MBF obtained in the step S4.
Specifically, the gray scale map of the output myocardial blood flow velocity MBF is converted into a color code parameter map, and then the load state myocardial blood flow velocity: resting myocardial blood flow rate to obtain coronary artery blood flow reserve fraction
The method further comprises the following steps: a preprocessing module for implementing the step S1; a DenseNet convolutional neural network module to implement the step S2; a GRU module for implementing the step S3; an upsampling module to implement said step S4; and communication connection is established among the preprocessing module, the DenseNet convolutional neural network module, the GRU module and the up-sampling module and is used for data transmission.
It should be noted that:
in one embodiment, the input to the pre-processing module is a 4D image of PET in JPEG/PNG format. Outputting normalized tomographic images of 128 × 128 × n × t (n = number of image layers of z-axis; t = total number of frames of dynamic acquisition time sequence, which can be set according to different PET acquisition sequences), for further data enhancement of image diversity and reduction of overfitting phenomenon, and specifically comprising the steps of:
step a, the resampled three-dimensional volume of a training data set is 128 multiplied by n pixels;
and b, using Gaussian filtering smoothing and anisotropic diffusion filtering preprocessing treatment, thereby reducing image noise in the training data and simultaneously keeping the detail and edge important information of the image.
Step c, further enhancing data of the filtered image, randomly turning/rotating the y axis and the z axis at each angle, and finely adjusting the contrast; randomly moving the time sequence t by the time frame number of-5 to +5, wherein the total frame number is not changed;
the input of the DenseNet convolutional neural network module for extracting the spatial characteristics is a preprocessed and data-enhanced PET multi-plane three-dimensional image. The method aims to directly extract and down-sample three-dimensional space features on each time frame to obtain high-order feature input of the next step, and specifically comprises the following steps:
step d, the preprocessed image is processed into patch images for each time frame image according to time sequence to form a plurality of 32 multiplied by n three-dimensional patch samples, and the first frame (t) is arranged according to the default frame number 1 ) To the last frame (t) N ) As a synchronization information channel input;
step e, using a DenseNet-Block convolutional neural network, processing all three-dimensional patches on each frame image layer by a convolution algorithm core with the filtering size of 3 × 3 × 3, wherein the algorithm of each layer of CNN comprises a ReLU activation function layer (f (x) = max (0,x)), a normalization layer and the output of space volume characteristics; further, the three-dimensional DenseNet extracts the spatial features of a time frame. And deriving learned featuremaps from each convolution layer to form input of the next layer, and gradually outputting the learned featuremaps after the learned featuremaps are advanced forwards.
Further, the GRU module recursively updates the existing time sequence for the extracted image spatial features, so that each layer is in a form of cyclic update and forward output as an input of a next unit. Specifically, as shown in fig. 3, the input of the module is the spatial feature extracted by the three-dimensional DenseNet in the previous step, and the purpose is to extract the time sequence relation feature of the dynamic PET through the GRU convolutional neural network algorithm, and the GRU is used to perform their time sequence relation processing on the spatial feature output in the step 2, that is, the convolutional layer of the previous frame is output (h) on the existing convolutional neural unit t-1 ) Performing memory/forgetting, parallel combining existing input X t The existing time sequence is updated recursively through the activation functions sigmoid (sigma), tanh and weight distribution, so that each layer can be circulatedUpdating and outputting the form forward, and further serving as the input of the next unit, further, the GUR processes the time sequence relationship, each a is a unit in the time sequence relationship, and the GUR considers the output of the previous unit and the influence of the existing input on the output of the existing unit.
Upsampling module and final output: and inputting the dynamic time sequence characteristics extracted by the GUR in the last step. The method aims to restore the resolution and the details of an image and output the myocardial blood flow velocity MBF through an up-sampling module, namely, a 3D quantitative parameter graph which can be analyzed is formed, and further, the up-sampling module performs up-sampling on the three-dimensional time sequence characteristics of the GRU by using multilayer deconvolution calculation;
it should be further noted that, in the present invention, down-sampling: when CNN is convolved, the step size of the convolution kernel is larger than 1, or when the step size is larger than 1 during pooling, so that the output image is smaller than the input image, which is equivalent to down-sampling.
And (3) upsampling: the deconvolution operation of CNN is to make the output image larger than the input image.
For better understanding of the present invention, which is only used as an example and not limiting the present invention, before the algorithm training begins, all image data will be divided into training set and validation set, and the general ratio is training set: validation set =70%:30 percent. The target or golden standard image (Ground truth = GT, labeled h) of the training set is 15 The input of the 3D planar image of myocardial blood flow of O-PET is 15 3D planar image of myocardial blood flow of O-PET. Through the calculation of the multiple deep convolution neural network DenseNet and GRU, a predicted MBF parameter map (marked as h') is output. The training aims at finding the minimum of the loss function (L = h-h') between the GT image and the predicted image. In addition to this, the algorithm also incorporates L1 regularization to the loss function (L1 regularization) to prevent overfitting. The loss function here is taken as cross entropy (CN). In the training set and the verification set, a back-propagation (BP) method and an Adam optimizer (Adam optimization) are used for carrying out gradient descent feedback and updating the weights of the optimization convolution networkThe parameters are re-evaluated to minimize the loss function L. Initial learning rate of 10 -3 When the loss function of the data on the verification set is not improved, the initial learning rate is gradually reduced after a plurality of training rounds (the attenuation rate is 10) -4 Max 100 rounds) to adjust learning time and performance optimization in time. All code is developed and implemented in python3.0 or above environment using Keras framework, and the hardware is generally a plurality of Nvidia GeForce Titan X image processing graphics cards.
For further understanding of the present invention coronary flow reserve fraction,
for example, when injecting a drug (F18-FDG) into a patient, a myocardial blood flow rate quantitative parameter Map (MBF) of the patient may be obtained according to the method described above, where a resting state of MBF is obtained, for example, at about 10min intervals, the drug is injected again, where a loading state of MBF is obtained, and the loading state of MBF: and obtaining CFR by the resting MBF.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. A myocardial ischemia coronary blood flow reserve analysis method is characterized in that based on multiple deep convolutional neural networks DenseNet and GRU, the method comprises the following steps:
s1, preprocessing an input image to obtain a myocardial dynamic PET three-dimensional image;
s2, respectively extracting image space characteristics and performing downsampling on the image preprocessed in the step S1 by using a DenseNet network architecture of a multiple convolution neural network;
s3, extracting dynamic time sequence characteristics of the image space characteristics extracted in the step S2 by using GRU;
s4, forming an analyzed three-dimensional quantitative parameter map for the image space characteristics processed in the step S3 through an up-sampling DenseNet architecture to obtain a myocardial blood flow velocity MBF;
and S5, obtaining the coronary blood flow reserve fraction CFR according to the images of the PET in the resting state and the loading state of the myocardial function and the myocardial blood flow velocity MBF obtained in the step S4.
2. The method for analyzing myocardial ischemic coronary flow reserve according to claim 1, further comprising:
a preprocessing module for implementing the step S1;
a DenseNet convolutional neural network module to implement the step S2;
a GRU module to implement said step S3;
an upsampling module for implementing the step S4;
and communication connection is established among the preprocessing module, the DenseNet convolutional neural network module, the GRU module and the up-sampling module and is used for data transmission.
3. The method for analyzing myocardial ischemia coronary flow reserve according to claim 2, wherein the preprocessing module comprises the steps of:
step a, resampling a three-dimensional volume of a training data set to 128 multiplied by n pixels;
b, using Gaussian filtering smoothing and anisotropic diffusion filtering pretreatment;
step c, further enhancing data of the filtered image, randomly turning/rotating the y axis and the z axis at each angle, and finely adjusting the contrast; randomly moving the time sequence t by-5 to +5 time frame numbers, wherein the total frame number is unchanged;
wherein n = z-axis image layer number, t = dynamic acquisition time sequence total frame number, which can be set according to different PET acquisition sequences.
4. The method for analyzing myocardial ischemic coronary flow reserve according to claim 2, wherein the DenseNet convolutional neural network module comprises the steps of:
d, carrying out spot processing on each time frame image according to a time sequence on the preprocessed image to form a plurality of 32 multiplied by n three-dimensional spot samples, and arranging a first frame to a last frame according to a default frame number to be used as a synchronous information channel input;
and e, using a multilayer DenseNet-Block convolutional neural network, and processing all three-dimensional patches on each frame of image layer by using a convolution algorithm core with the filtering size of 3 multiplied by 3.
5. The method of claim 2, wherein the GRU module recursively updates the existing time sequence for the extracted image spatial features such that each layer is cyclically updated and output onward as input to a next unit.
6. The method of analyzing myocardial ischemic coronary flow reserve of claim 2, wherein the upsampling module upsamples three-dimensional timing features of the GRUs using a multi-layer deconvolution calculation.
7. The method for analyzing myocardial ischemia coronary flow reserve according to claim 1, wherein in step S5, the gray scale map of the output myocardial blood flow velocity MBF is converted into a color code parameter map, and then the load-state myocardial blood flow velocity: and obtaining the coronary blood flow reserve fraction by the resting myocardial blood flow rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910507174.XA CN110136810B (en) | 2019-06-12 | 2019-06-12 | Analysis method of myocardial ischemia coronary blood flow reserve |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910507174.XA CN110136810B (en) | 2019-06-12 | 2019-06-12 | Analysis method of myocardial ischemia coronary blood flow reserve |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110136810A CN110136810A (en) | 2019-08-16 |
| CN110136810B true CN110136810B (en) | 2023-04-07 |
Family
ID=67581317
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910507174.XA Active CN110136810B (en) | 2019-06-12 | 2019-06-12 | Analysis method of myocardial ischemia coronary blood flow reserve |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110136810B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110432886A (en) * | 2019-08-22 | 2019-11-12 | 苏州润心医疗器械有限公司 | Obtain the mean blood flow in coronary artery exit, the methods, devices and systems of flow velocity |
| CN112001486B (en) * | 2020-08-28 | 2022-07-22 | 河北工业大学 | Load decomposition method based on deep learning |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106023202A (en) * | 2016-05-20 | 2016-10-12 | 苏州润心医疗科技有限公司 | Coronary artery fractional flow reserve calculation method based on heart CT image |
| CN106327487A (en) * | 2016-08-18 | 2017-01-11 | 苏州润心医疗科技有限公司 | Coronary artery blood flow reserve fraction calculation method based on X ray coronary artery angiographic image |
| CN106408562A (en) * | 2016-09-22 | 2017-02-15 | 华南理工大学 | Fundus image retinal vessel segmentation method and system based on deep learning |
| CN108511075A (en) * | 2018-03-29 | 2018-09-07 | 向建平 | A kind of non-intrusion type obtains the method and system of blood flow reserve score |
| CN109326354A (en) * | 2018-11-09 | 2019-02-12 | 深圳市孙逸仙心血管医院（深圳市心血管病研究所） | Based on ANN blood flow reserve Score on Prediction method, apparatus, equipment and medium |
| CN109344815A (en) * | 2018-12-13 | 2019-02-15 | 深源恒际科技有限公司 | A kind of file and picture classification method |
| CN109754306A (en) * | 2018-11-13 | 2019-05-14 | 北京码牛科技有限公司 | Information processing method and device, electronic equipment and computer-readable medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9700219B2 (en) * | 2013-10-17 | 2017-07-11 | Siemens Healthcare Gmbh | Method and system for machine learning based assessment of fractional flow reserve |
-
2019
- 2019-06-12 CN CN201910507174.XA patent/CN110136810B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106023202A (en) * | 2016-05-20 | 2016-10-12 | 苏州润心医疗科技有限公司 | Coronary artery fractional flow reserve calculation method based on heart CT image |
| CN106327487A (en) * | 2016-08-18 | 2017-01-11 | 苏州润心医疗科技有限公司 | Coronary artery blood flow reserve fraction calculation method based on X ray coronary artery angiographic image |
| CN106408562A (en) * | 2016-09-22 | 2017-02-15 | 华南理工大学 | Fundus image retinal vessel segmentation method and system based on deep learning |
| CN108511075A (en) * | 2018-03-29 | 2018-09-07 | 向建平 | A kind of non-intrusion type obtains the method and system of blood flow reserve score |
| CN109326354A (en) * | 2018-11-09 | 2019-02-12 | 深圳市孙逸仙心血管医院（深圳市心血管病研究所） | Based on ANN blood flow reserve Score on Prediction method, apparatus, equipment and medium |
| CN109754306A (en) * | 2018-11-13 | 2019-05-14 | 北京码牛科技有限公司 | Information processing method and device, electronic equipment and computer-readable medium |
| CN109344815A (en) * | 2018-12-13 | 2019-02-15 | 深源恒际科技有限公司 | A kind of file and picture classification method |
Non-Patent Citations (2)
| Title |
|---|
| 基于FCM聚类算法与自适应阈值的冠脉钙化分割与量化;赵聪等;《中国生物医学工程学报》;20171020(第05期);全文 * |
| 深度学习在图像识别中的应用研究综述;郑远攀等;《计算机工程与应用》;20190419(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110136810A (en) | 2019-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ye et al. | Multi-depth fusion network for whole-heart CT image segmentation | |
| CN109493347B (en) | Method and system for segmenting sparsely distributed objects in an image | |
| CN108198184B (en) | Method and system for vessel segmentation in contrast images | |
| CN109035255A (en) | A kind of sandwich aorta segmentation method in the CT image based on convolutional neural networks | |
| Zhang et al. | LU-NET: An improved U-Net for ventricular segmentation | |
| CN109544457A (en) | Image super-resolution method, storage medium and terminal based on fine and close link neural network | |
| CN111815574A (en) | Coarse set neural network method for fundus retina blood vessel image segmentation | |
| CN114581662B (en) | Brain tumor image segmentation method, system, device and storage medium | |
| CN110136810B (en) | Analysis method of myocardial ischemia coronary blood flow reserve | |
| US11816870B2 (en) | Image processing method and device, neural network and training method thereof, storage medium | |
| CN112150476A (en) | Coronary artery sequence vessel segmentation method based on space-time discriminant feature learning | |
| CN111583262A (en) | Blood vessel segmentation method and system | |
| CN109919915A (en) | Retinal fundus images abnormal area detection method and equipment based on deep learning | |
| CN115564649B (en) | Image super-resolution reconstruction method, device and equipment | |
| CN111798377A (en) | Magnetic resonance image super-resolution reconstruction method based on multi-resolution learning strategy | |
| CN114723698A (en) | Cerebrovascular image segmentation method based on multi-scale attention network | |
| CN111179269A (en) | PET image segmentation method based on multi-view and 3-dimensional convolution fusion strategy | |
| CN115100093A (en) | Medical image fusion method based on gradient filtering | |
| CN114897780A (en) | MIP sequence-based mesenteric artery blood vessel reconstruction method | |
| Ma et al. | Back-projection-based progressive growing generative adversarial network for single image super-resolution | |
| CN114581434A (en) | Pathological image processing method based on deep learning segmentation model and electronic equipment | |
| CN110276748B (en) | Method for analyzing blood flow velocity and fractional flow reserve of myocardial ischemia area | |
| Yang et al. | AMF-NET: Attention-aware multi-scale fusion network for retinal vessel segmentation | |
| CN112634308B (en) | Nasopharyngeal carcinoma target area and organ-at-risk delineating method based on different receptive fields | |
| Wang et al. | 3D dense convolutional neural network for fast and accurate single MR image super-resolution |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |