CN110428417A - Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques - Google Patents
Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques Download PDFInfo
- Publication number
- CN110428417A CN110428417A CN201910744483.9A CN201910744483A CN110428417A CN 110428417 A CN110428417 A CN 110428417A CN 201910744483 A CN201910744483 A CN 201910744483A CN 110428417 A CN110428417 A CN 110428417A
- Authority
- CN
- China
- Prior art keywords
- image
- information
- ultrasound
- training
- ultrasonic
- 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
- G06T2207/10132—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
- G06T2207/20—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)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The present invention relates to ultrasound image processing technology fields, specifically disclose a kind of property method of discrimination of carotid plaques, it include: to obtain at least two groups ultrasound information that same clients include carotid plaques, every group of ultrasound information includes clients based on any one obtained ultrasonic video or ultrasound image in a variety of ultrasonic examination modes, and the ultrasonic examination mode of at least two groups ultrasound information is different；By first nerves Network Recognition at least two groups ultrasound information, at least two groups vessel information is obtained, every group of vessel information includes blood vessel video or blood-vessel image；The characteristics of image at least two groups vessel information is extracted by Feature Selection Model；According to the characteristics of image in nervus opticus network and at least two groups vessel information, identification obtains the property class of patch in blood vessel.The invention also discloses a kind of storage medium and Vltrasonic devices.The property method of discrimination of carotid plaques provided by the invention can be realized the property comprehensive assessment to carotid plaques, improve the accuracy for differentiating result.
Description
Technical field
The present invention relates to the property method of discrimination of ultrasound image processing technology field more particularly to a kind of carotid plaques,
The storage medium of the program instruction of property method of discrimination of carotid plaques and surpassing including the storage medium are executed for storing
Acoustic device.
Background technique
Carotid plaques stability and vulnerability depend on the structure of its interior tissue, lipid necrotic cores size, patch
Internal haemorrhage, calcification, thickness and integrality of fibrous cap etc., the conventional inspection method based on ultrasound includes two-dimensional ultrasound, colored
Blood flow patterns, capilary imaging technique, contrast-enhanced ultrasound technique, the methods of ultrasonic elastograph imaging go to analyze its spot from all angles
The property of block, doctor, which is generally comprehensively considered by a variety of methods, stablizes rapid wear assessment for the patch of patient.
Summary of the invention
The present invention provides a kind of property method of discrimination of carotid plaques, for storing the property for executing carotid plaques
The storage medium of the program instruction of method of discrimination and Vltrasonic device including the storage medium solve present in the relevant technologies such as
What realizes the problem of carrying out comprehensive assessment to the property of carotid plaques.
As one aspect of the present invention, a kind of property method of discrimination of carotid plaques is provided, wherein the arteria carotis
The property method of discrimination of patch includes:
At least two groups ultrasound information that same clients include carotid plaques is obtained, every group of ultrasound information includes clients
Based on any one obtained ultrasonic video or ultrasound image in a variety of ultrasonic examination modes, and at least two groups ultrasound information
Ultrasonic examination mode is different；
By at least two groups ultrasound information described in first nerves Network Recognition, at least two groups vessel information, every group of institute are obtained
Stating vessel information includes blood vessel video or blood-vessel image, and the blood vessel video includes the blood vessel location for having patch in ultrasonic video
The image stream in domain, the blood-vessel image include the picture having where the blood vessel of patch in ultrasound image；
The characteristics of image in at least two groups vessel information is extracted by Feature Selection Model；
According to the characteristics of image in nervus opticus network and at least two groups vessel information, identification obtains patch in blood vessel
Property class.
Further, the property method of discrimination of the carotid plaques further include:
Training sample is obtained, the training sample includes that the n group training ultrasound information of m clients and every group of training surpass
The markup information of acoustic intelligence, and each clients correspond at least two groups training ultrasound information, every group of trained ultrasound information includes instruction
Practice ultrasonic video or training ultrasound image, the markup information include the first markup information and the second markup information, described first
Markup information is for marking the position having where the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image, institute
The second markup information, which is stated, for marking the property class of patch in the trained ultrasonic video or the trained ultrasound image, n is
Integer more than or equal to 2, m are the integer more than or equal to 1；
According to n group training ultrasound information and first markup information training the first initialization network, institute is obtained
State first nerves network；
According to n group training ultrasound information and second markup information training the second initialization network, institute is obtained
State nervus opticus network.
Further, the ultrasonic examination mode is two-dimensional ultrasound, capilary imaging technique, colorful blood, elastogram
Or ultrasonic contrast.
As another aspect of the present invention, a kind of property method of discrimination of carotid plaques is provided, wherein the neck is dynamic
The property method of discrimination of arteries and veins patch includes:
Training sample is obtained, the training sample includes that the n group training ultrasound information of m clients and every group of training surpass
The markup information of acoustic intelligence, every group of trained ultrasound information include training ultrasonic video or training ultrasound image, and each clients
Corresponding at least two groups training ultrasound information, the markup information include the first markup information and the second markup information, and described first
Markup information is for marking the position having where the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image, institute
The second markup information, which is stated, for marking the property class of patch in the trained ultrasonic video or the trained ultrasound image, n is
Integer more than or equal to 2, m are the integer more than or equal to 1；
According to n group training ultrasound information and first markup information training the first initialization network, the is obtained
One neural network, the first nerves network have patch in the trained ultrasonic video or the trained ultrasound image for identification
Blood vessel where position；
The characteristics of image in the n group training ultrasound information is extracted, characteristic aggregation, root are carried out to the characteristics of image extracted
According to the characteristics of image and second markup information training the second initialization network after characteristic aggregation, nervus opticus net is obtained
Network, the property class of nervus opticus network patch for identification.
Further, if in n group ultrasound training information including the ultrasonic training information for meeting preset condition, described
According to n group training ultrasound information and first markup information training the first initialization network, comprising:
By in the ultrasonic training information for meeting the preset condition two-dimensional image information and other types image information into
Row fusion；
According to fused information and first markup information training the first initialization network.
Further, the two-dimensional image information by the ultrasonic training information for meeting the preset condition and other classes
Type image information is merged, comprising:
Respectively the gray value of position in the two-dimensional image information and other types image information is determined as melting
The value of image after conjunction in the channel B of corresponding position；
The gray value of position in the two-dimensional image information is determined as fused image in the G of corresponding position
The value in channel；
Default value is determined as value of the fused image at the channel R.
Further, the described pair of characteristics of image extracted carries out characteristic aggregation, according to the characteristics of image after characteristic aggregation
And the second initialization of the second markup information training network, obtain nervus opticus network, comprising:
The characteristics of image extracted is subjected to merging features, obtains spliced characteristics of image；
According to the spliced characteristics of image and second markup information training the second initialization network, the is obtained
Two neural networks.
Further, characteristic aggregation is carried out to the characteristics of image that extracts, according to after characteristic aggregation characteristics of image and
Second markup information training the second initialization network, before obtaining nervus opticus network, further includes:
Video image in the trained ultrasonic video is carried out image to convert to obtain the transformed video figure of described image
Picture, wherein described image transformation includes at least one of translation, overturning, rotation and flexible deformation；
The characteristics of image in the transformed video image of described image is extracted by Feature Selection Model.
As another aspect of the present invention, a kind of storage medium is provided, wherein be stored at least in the storage medium
One program instruction, at least one program instruction is for being loaded and being executed by processor to realize that neck as previously described moves
The property method of discrimination of arteries and veins patch.
As another aspect of the present invention, provide a kind of Vltrasonic device, wherein the Vltrasonic device include processor and
Memory, the memory include previously described storage medium, and the processor is for loading and executing the storage medium
On program instruction.
By property method of discrimination, storage medium and the Vltrasonic device of above-mentioned carotid plaques, using first nerves network
To under multi-modal ultrasonic video and/or ultrasound image identify, image characteristics extraction is carried out by Feature Selection Model, most
Characteristics of image is identified by nervus opticus network afterwards, obtains the differentiation of the property class of final blood vessel as a result, above-mentioned
The property method of discrimination of carotid plaques can effectively identify the property class of the blood vessel of patch, realize to arteria carotis
The qualitative analysis of blood vessel, to provide important foundation for analysis and diagnosis of the medical staff to carotid plaques.In addition, above-mentioned
The property method of discrimination of carotid plaques additionally use under multi-modal ultrasound image and/or ultrasonic video analyzed,
Comprehensive analysis, this property method of discrimination energy based on multi-modal carotid plaques are carried out for a variety of ultrasonic examination modes
The enough property for patch carries out comprehensive assessment, so as to improve the accuracy of analysis.
Detailed description of the invention
The drawings are intended to provide a further understanding of the invention, and constitutes part of specification, with following tool
Body embodiment is used to explain the present invention together, but is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the structural block diagram of the property judgement system of carotid plaques provided by the invention.
Fig. 2 is a kind of flow chart of embodiment of the property method of discrimination of carotid plaques provided by the invention.
Fig. 3 is the flow chart of the another embodiment of the property method of discrimination of carotid plaques provided by the invention.
Fig. 4 is the mark schematic diagram of the training ultrasonic video under ultrasonic contrast mode provided by the invention.
Fig. 5 is the mark schematic diagram of the training ultrasonic video under capilary imaging mode provided by the invention.
Fig. 6 is the mark schematic diagram of the training ultrasound image under two-dimensional ultrasound mode provided by the invention.
Fig. 7 is the mark schematic diagram of the training ultrasound image under colorful blood mode provided by the invention.
Fig. 8 is the mark schematic diagram of the training ultrasound image under elastogram mode provided by the invention.
Fig. 9 is the schematic diagram before image provided by the invention transformation.
Figure 10 is the transformed schematic diagram of image provided by the invention.
Figure 11 is the flow diagram that nervus opticus network provided by the invention carries out feature identification.
Figure 12 is that the fully-connected network of Mixed Weibull distribution provided by the invention realizes schematic diagram.
Specific embodiment
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase
Mutually combine.The present invention will be described in detail below with reference to the accompanying drawings and embodiments.
In order to make those skilled in the art more fully understand the present invention program, below in conjunction with attached in the embodiment of the present invention
Figure, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only this
The embodiment of a part is invented, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, should fall within the scope of the present invention.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way
Data be interchangeable under appropriate circumstances, so as to the embodiment of the present invention described herein.In addition, term " includes " and " tool
Have " and their any deformation, it is intended that cover it is non-exclusive include, for example, containing a series of steps or units
Process, method, system, product or equipment those of are not necessarily limited to be clearly listed step or unit, but may include without clear
Other step or units listing to Chu or intrinsic for these process, methods, product or equipment.
In one embodiment of the present of invention, a kind of judgement system of property class for differentiating carotid plaques is provided, such as
Shown in Fig. 1, which includes obtaining module 110, processor 120 and display 130.Obtain module 110 obtain it is same just
The person of examining includes at least two groups ultrasound information of carotid plaques, and processor 120 handles collected ultrasound information, analysis
The property class of patch is obtained, display 130 can show the property for the patch that the ultrasound information got and analysis obtain
Classification, wherein every group of ultrasound information includes clients based on any one obtained ultrasonic video in a variety of ultrasonic examination modes
Or ultrasound image, and the different a variety of ultrasonic examination modes of ultrasonic examination mode of at least two groups ultrasound information can be super for two dimension
Sound, capilary imaging technique, colorful blood, elastogram or ultrasonic contrast；Wherein, the ultrasound inspection of at least two groups ultrasound information
Mode difference is looked into it is to be understood that combining to obtain ultrasound information using different ultrasonic examination modes, such as can be using two dimension
The combination test mode of ultrasound and capilary imaging technique can also use the combined reviewing party of two-dimensional ultrasound and ultrasonic contrast
Formula can also can also be upper using the combined test mode of two-dimensional ultrasound, capilary imaging technique and ultrasonic contrast certainly
State 5 kinds of combined test mode of citing.
It should be understood that when using any one or more reviewing party in two-dimensional ultrasound, colorful blood and elastogram
What is obtained when formula is ultrasound image, and what is obtained when using the test mode of capilary imaging technique and/or ultrasonic contrast is super
Sound video.
As shown in Figure 1, the acquisition module 110 of the present embodiment is supersonic imaging apparatus, i.e., obtained by supersonic imaging apparatus
Ultrasound image or video.As shown in Figure 1, supersonic imaging apparatus includes at least energy converter 101, ultrasonic host 102, input unit
103, control unit 104 and memory 105.Supersonic imaging apparatus may include display screen (not marking in figure), and ultrasonic imaging is set
Standby display screen can be the display 130 of judgement system.The ultrasonic wave for transmitting and receiving of energy converter 101, energy converter 101 by
Exomonental excitation emits ultrasound to destination organization (for example, human body or the intracorporal organ of animal, tissue, blood vessel etc.)
Wave receives the ultrasonic echo from the reflected information with destination organization in target area after certain time-delay, and this is surpassed
Sound echo is converted to electric signal again, to obtain ultrasound image or video.Energy converter 101 can pass through wired or wireless side
Formula is connected to ultrasonic host 102.
Input unit 103 is used to input the control instruction of operator.Input unit 103 can be keyboard, tracking ball, mouse
At least one of mark, touch panel, handle, driver plate, control stick and floor push.Input unit can also input non-contact
Type signal, such as sound, gesture, sight or brain wave signal.
Control unit 104 at least can control focus information, drive frequency information, driving voltage information and imaging pattern
Equal scanning informations.The difference of the imaging pattern according to needed for user of control unit 104 carries out different processing to signal, obtains not
With the ultrasound image data of mode, different moulds then are formed through processing such as log-compressed, dynamic range adjustment, digital scan conversions
The ultrasound image of formula, such as B image, C image, D image, doppler flow inaging image, the elastic image etc. comprising tissue elasticity characteristic
Deng or other kinds of two-dimensional ultrasonic image or three-dimensional ultrasound pattern.
Display 130 for show in ultrasound image data, parameter, ultrasound image or video perineural type and
The information such as multidate information.Display 130 can be touch-screen display.Certainly, ultrasonic diagnostic equipment can also be single by input
The port of member 103 connects another display, realizes Double-Screen Display System.In addition, the display in the present embodiment does not limit number
Amount.The ultrasound image data (ultrasound image) of display can be shown on a display, can also be simultaneously displayed on multiple
On display, naturally it is also possible to be the part difference simultaneous display by ultrasound image on multiple displays, in this present embodiment
Also with no restriction.In addition, display 130 is also provided to figure circle that user carries out human-computer interaction while showing ultrasound image
One or more controlled devices are arranged on graphical interfaces, user is supplied to and is input operation instruction using human-computer interaction device for face
These controlled devices are controlled, thereby executing corresponding control operation.Such as projection, VR glasses, it can also be in certain display
Include input unit, such as projector's VR glasses that the display screen of touch input, induction act.It can be with using human-computer interaction device
The icon shown to display 130 operates, for executing specific function.
In one embodiment, obtaining module 110 is applause ultrasound, and energy converter, the display etc. of applause ultrasound are integrated in can
In the shell held for operator's hand.
Neural network model or unit of the invention includes (either comprising or have) other elements and those elements.
Term " module " as used in the present invention means but is not limited to the software or hardware component of execution particular task, and such as scene can
Program gate array (FPGA) or specific integrated circuit (ASIC) or processor, such as CPU, GPU.Module can advantageously be configured
It is executed on the one or more processors to reside in addressable storage medium and being configured to.Therefore, as an example, module
May include component (such as component software, object oriented software component, class component and task component), process, function, attribute,
Process, subroutine, program code segments, driver, firmware, microcode, circuit, data, database, data structure, table, number
Group and variable.The functionality provided in the module can be combined into less component and module or be further separated into additional
Component and module.
In order to realize the qualitative analysis to the property of carotid plaques, by the present invention in that with two neural fusions,
And ultrasonic video or ultrasound image under a variety of ultrasonic examination modes is combined to carry out comprehensive analysis, specifically, first nerve net
Network is capable of detecting when the vessel position of the patch in the ultrasonic video and/or ultrasound image under current each ultrasonic examination mode,
Second neural network can identify the blood vessel containing patch, thus identify the property class of blood vessel be rapid wear or
Stablize.
A kind of property method of discrimination of carotid plaques is provided in the present embodiment, and Fig. 2 is according to embodiments of the present invention
The flow chart of the property method of discrimination of the carotid plaques of offer, as shown in Figure 2, comprising:
S210, at least two groups ultrasound information that same clients include carotid plaques is obtained, every group of ultrasound information includes
Clients are based on any one obtained ultrasonic video or ultrasound image in a variety of ultrasonic examination modes, and at least two groups are ultrasonic
The ultrasonic examination mode of information is different；
This step can be the ultrasonic video and/or ultrasound image acquired by energy converter, be also possible to the outside obtained
The ultrasonic video and/or ultrasound image that equipment is sent.For example, ultrasonic device or computer etc. are collecting ultrasonic view by energy converter
After frequency and/or ultrasound image, ultrasonic video and/or ultrasound image are sent to analytical equipment, analytical equipment receive it is super
Step S210~S240 is executed after sound video and/or ultrasound image, details are not described herein.
It should be noted that the ultrasonic examination mode used in the present invention can specifically include two-dimensional ultrasound, capilary at
As at least two in technology, colorful blood, elastogram and ultrasonic contrast, to include two-dimensional ultrasound, micro- blood in the present embodiment
It is illustrated for whole in pipe imaging technique, colorful blood, elastogram and ultrasonic contrast.For different ultrasonic examination
Mode emphasis is different, ultrasonic contrast and capilary imaging technique be mainly based upon dynamic mode observe it is new inside patch
The situation of change of angiogenic, and two-dimensional ultrasound, colorful blood and elastogram are the profiles that patch is observed based on static mode
And therefore property for ultrasonic contrast and capilary imaging technique both test modes is carried out in the form of ultrasonic video
Acquisition and mark, two-dimensional ultrasound, colorful blood and elastogram are acquired and are marked in the form of ultrasound image.
S220, pass through at least two groups ultrasound information described in first nerves Network Recognition, obtain at least two groups vessel information, often
The group vessel information includes blood vessel video or blood-vessel image, and the blood vessel video includes the blood vessel institute for having patch in ultrasonic video
Image stream in region, the blood-vessel image include the picture having where the blood vessel of patch in ultrasound image；
Specifically, at least two groups ultrasound information described in the first nerves Network Recognition, in order to determine ultrasonic video and/
Or the position of the blood vessel for having patch in ultrasound image.Since the ultrasonic video includes multiple image, first nerves net
Network can be from identifying in ultrasonic video with the image where the blood vessel of patch, here it should be understood that since ultrasound regards
Frequency is dynamic video image, so first nerves network finally identifies that obtain is dynamic image stream for ultrasonic video.
Since the ultrasound image includes plurality of pictures, first nerves network can be identified from ultrasound image
With the picture where the blood vessel of patch.Here it should be understood that since ultrasound image is static image, so the first mind
Finally identify that is obtained is static picture for ultrasound image through network.
S230, the characteristics of image in at least two groups vessel information is extracted by Feature Selection Model；
Specifically, video processing is carried out to the ultrasonic video, or image procossing is carried out to ultrasound image, extraction is provided
There is the characteristics of image of the blood vessel of patch.
S240, according to the characteristics of image in nervus opticus network and at least two groups vessel information, identification obtains blood vessel
The property class of middle patch.
Specifically, the nervus opticus network can be in conjunction with the characteristics of image in at least two groups vessel information being previously obtained
Described image feature is identified, so that the property class that judgement has the blood vessel of patch is rapid wear or stabilization.
The property method of discrimination of carotid plaques in through the foregoing embodiment, using first nerves network to a variety of ultrasounds
Ultrasonic video and/or ultrasound image under test mode are identified, carry out image characteristics extraction by Feature Selection Model, most
Characteristics of image is identified by nervus opticus network afterwards, obtains the differentiation of the property class of final blood vessel as a result, above-mentioned
The property method of discrimination of carotid plaques can effectively identify the property class of the blood vessel of patch, realize to arteria carotis
The qualitative analysis of blood vessel, to provide important foundation for analysis and diagnosis of the medical staff to carotid plaques.In addition, above-mentioned
The property method of discrimination of carotid plaques additionally use under multi-modal ultrasound image and/or ultrasonic video analyzed,
Comprehensive analysis, this property method of discrimination energy based on multi-modal carotid plaques are carried out for a variety of ultrasonic examination modes
The enough property for patch carries out comprehensive assessment, so as to improve the accuracy of analysis.
Wherein, first nerves network and nervus opticus network are the networks that preparatory training obtains, and trained step includes:
Training sample is obtained, the training sample includes that the n group training ultrasound information of m clients and every group of training surpass
The markup information of acoustic intelligence, and each clients correspond at least two groups training ultrasound information, every group of trained ultrasound information includes instruction
Practice ultrasonic video or training ultrasound image, the markup information include the first markup information and the second markup information, described first
Markup information is for marking the position having where the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image, institute
The second markup information, which is stated, for marking the property class of patch in the trained ultrasonic video or the trained ultrasound image, n is
Integer more than or equal to 2, m are the integer more than or equal to 1；
According to n group training ultrasound information and first markup information training the first initialization network, institute is obtained
State first nerves network；
According to n group training ultrasound information and second markup information training the second initialization network, institute is obtained
State nervus opticus network.
It should be understood that the trained ultrasound information of n of m clients can be history acquisition, checked for example, obtaining
The ultrasonic video and/or ultrasound image that department acquires in history 3 months for another example obtain laboratory and adopt in history 1 year
The ultrasonic video and/or ultrasound image of collection.Wherein, in order to realize in ultrasonic video or ultrasound image with the blood vessel of patch
Image recognition, then the position where needing to have the blood vessel of patch on mark in training ultrasonic video and/or training ultrasound image,
In order to realize the identification of the property class to the patch in ultrasonic video or ultrasound image, then need training ultrasonic video and/
Or have the property class of patch in training ultrasound image on mark, i.e., delicate patch and stable patch are marked respectively
Note, to be trained.
It is to be further understood that according at the beginning of n group training ultrasound information and first markup information training first
Beginningization network, obtaining the first nerves network can specifically include following several situations:
(1) when ultrasonic examination mode is static mode, for example, with two-dimensional ultrasound, colorful blood and three kinds of elastogram
What mode was realized, then according to n group training ultrasound image and first markup information training the first initialization network, obtain
To the first nerves network, the first markup information described herein has patch particularly for marking in the trained ultrasound image
Blood vessel where position；
(2) when ultrasonic examination mode is dynamical fashion, for example, with ultrasonic contrast and capilary imaging technique two ways
It realizes, then according to n group training ultrasonic video and first markup information training the first initialization network, obtains institute
First nerves network is stated, the first markup information described herein has the blood of patch particularly for marking in the trained ultrasonic video
Position where pipe；
(3) when ultrasound detection mode is static mode combination dynamical fashion, for example, skill is imaged with two-dimensional ultrasound, capilary
Five kinds of art, colorful blood, elastogram and ultrasonic contrast modes are realized, then according to x group training ultrasonic video and n-x group
Training ultrasound image and first markup information training the first initialization network, obtain the first nerves network, herein
First markup information has the blood of patch particularly for marking in the trained ultrasonic video and the trained ultrasound image
Position where pipe.
Similarly, for according to n group training ultrasound information and second markup information training the second initialization net
Network, obtains the nervus opticus network, specifically also may include above-mentioned three kinds of situations, details are not described herein again.
As specifically embodiment, when acquiring for ultrasonic video, per second 10 can be sampled out according to the frame per second of 10fps
Image is opened, for example, can respectively acquire 100 frames (10s) for the ultrasonic video of ultrasonic contrast and capilary imaging technique.For two
The ultrasound image of dimension ultrasound, colorful blood and elastogram can respectively acquire 3 frames.It should be understood that being examined for each ultrasound
The quantity that the ultrasonic video and ultrasound image looked under mode acquire respectively can be configured according to demand, herein without limitation.
The embodiment of the property method of discrimination of above-mentioned carotid plaques is to be based on directly carrying out property on ultrasonic device
Differentiate realization；Certainly the property differentiation to patch can also be realized on background server.
As another embodiment of the present invention, a kind of property method of discrimination of carotid plaques is provided, wherein such as Fig. 3
Shown, the property method of discrimination of the carotid plaques includes:
S310, training sample is obtained, the training sample includes that the n group of m clients trains ultrasound information and every group
The markup information of training ultrasound information, every group of trained ultrasound information include training ultrasonic video or training ultrasound image, and each
Clients correspond at least two groups training ultrasound information, and the markup information includes the first markup information and the second markup information, institute
The first markup information is stated for where marking and having the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image
Position, second markup information are used to mark the property class of patch in the trained ultrasonic video or the trained ultrasound image
Not, n is the integer more than or equal to 2, and m is the integer more than or equal to 1；
S320, network is initialized according to n group training ultrasound information and first markup information training first, obtained
To first nerves network, the first nerves network has in the trained ultrasonic video or the trained ultrasound image for identification
Position where the blood vessel of patch；
Characteristics of image in S330, the extraction n group training ultrasound information, it is poly- to carry out feature to the characteristics of image extracted
It closes, according to the characteristics of image and second markup information training the second initialization network after characteristic aggregation, obtains the second mind
Through network, the property class of nervus opticus network patch for identification.
The property method of discrimination of carotid plaques in through the foregoing embodiment obtains training sample and obtains first nerves net
Corresponding neural network is respectively trained by markup information in markup information needed for network and nervus opticus network, what training obtained
First nerves network can identify the position where having the blood vessel of patch, and the nervus opticus network that training obtains can be realized to spot
The identification of the property class of block can be realized the qualitative analysis to carotid artery vascular by the two neural networks, so as to
Important foundation is provided for analysis and diagnosis of the medical staff to carotid plaques.In addition, being adopted during obtaining training sample
With under multi-modal training ultrasound image and/or training ultrasonic video analyzed, that is, be directed to a variety of ultrasonic examination sides
Formula carries out comprehensive analysis, and this property method of discrimination based on multi-modal carotid plaques can carry out the property of patch
Comprehensive assessment, so as to improve the accuracy of analysis.
It should be understood that the trained ultrasonic video and/or training ultrasound image are the carotid plaques by acquisition
Ultrasonic video and/or ultrasound image handled after obtain.And to the mark of training ultrasonic video and/or training ultrasound image
Note is mostly to be carried out by marking software, i.e., handles to ultrasonic video and/or ultrasound image, removes non-sensitive part, only
Retaining effective ultrasound image part, feeding marking software is labeled after being decoded into image to ultrasonic video, and for super
Acoustic image can be sent directly into marking software and be labeled.
It should be noted that the ultrasonic examination mode used in the present invention can specifically include two-dimensional ultrasound, capilary at
As at least two in technology, colorful blood, elastogram and ultrasonic contrast, to include two-dimensional ultrasound, micro- blood in the present embodiment
It is illustrated for whole in pipe imaging technique, colorful blood, elastogram and ultrasonic contrast.For different ultrasonic examination
Mode emphasis is different, ultrasonic contrast and capilary imaging technique be mainly based upon dynamic mode observe it is new inside patch
The situation of change of angiogenic, and two-dimensional ultrasound, colorful blood and elastogram are the profiles that patch is observed based on static mode
And therefore property for ultrasonic contrast and capilary imaging technique both test modes is carried out in the form of ultrasonic video
Acquisition and mark, two-dimensional ultrasound, colorful blood and elastogram are acquired and are marked in the form of ultrasound image.
Specifically, when being labeled to training ultrasonic video, it can use 10fps frame per second is therefrom per second to sample out 10
Image, sketched the contours of with rectangle frame include patch vessel profile, and the classification of the patch is marked out, as Fig. 4 makes for ultrasound
The mark schematic diagram of training ultrasonic video under shadow mode, Fig. 5 are the training ultrasonic video under capilary imaging technique mode
Mark schematic diagram.When being labeled to training ultrasound image, packet directly can be sketched the contours of with rectangle frame for the picture of acquisition
Vessel profile containing patch, and the classification of the patch is marked out, being the training ultrasound image under two-dimensional ultrasound mode such as Fig. 6
Schematic diagram is marked, Fig. 7 is the mark schematic diagram of the training ultrasound image under colorful blood mode, and Fig. 8 is under elastogram mode
Training ultrasound image mark schematic diagram.
As a kind of specifically embodiment of the present embodiment, if including meeting to preset in n group ultrasound training information
The ultrasonic training information of condition, it is described according to the first initialization of n group training ultrasound information and first markup information training
Network, comprising:
By in the ultrasonic training information for meeting preset condition two-dimensional image information and other types image information carry out
Fusion；
According to fused image information and first markup information training the first initialization network.
Further specifically, described merge includes:
Respectively the gray value of position in the two-dimensional image information and other types image information is determined as melting
The value of image after conjunction in the channel B of corresponding position；
The gray value of position in two dimensional image is determined as fused image in the G channel of corresponding position
Value；
Default value is determined as value of the fused image at the channel R.
For example, setting the random number between 0 to 1 for the weight parameter of the first initialization network, and use video image
And/or the above-mentioned first initialization network of image parameter training of picture, deconditioning after iteration to loss (loss) is minimum.
It should be understood that in the present embodiment with two-dimensional ultrasound, capilary imaging technique, colorful blood, elastogram and
It is illustrated for whole in ultrasonic contrast, then the summation of the training ultrasonic video that training sample includes and training ultrasound image
It is n.If it is super that training sample only includes n training in a manner of two-dimensional ultrasound, colorful blood and this static state of elastogram
Acoustic image, if by ultrasonic contrast and capilary imaging technique it is this it is dynamic in a manner of, training sample only include n training ultrasound
Video.
As another specifically embodiment of the present embodiment,
It is described to merge the characteristics of image extracted, according to the characteristics of image and second markup information after characteristic aggregation
Training the second initialization network, obtains nervus opticus network, comprising:
The n group characteristics of image is subjected to merging features, obtains spliced characteristics of image；
According to the spliced characteristics of image and second markup information training the second initialization network, the is obtained
Two neural networks.
Further specifically,
Extract the characteristics of image in the n group training ultrasound information, the characteristics of image that characteristic aggregation extracts, according to feature
Characteristics of image and second markup information training the second initialization network after polymerization, before obtaining nervus opticus network,
Further include:
Video image in the trained ultrasonic video is carried out image to convert to obtain the transformed video figure of described image
Picture, wherein described image transformation includes at least one of translation, overturning, rotation and flexible deformation；
The characteristics of image in the transformed video image of described image is extracted by Feature Selection Model.
It should be noted that when including training ultrasonic video in training sample, by the beginning of training ultrasonic video training second
When beginningization is netted, need to carry out training ultrasonic video enhancing image training.The present embodiment specifically using image transformation by the way of into
The enhancing of row image data.
Further specifically, the video image progress image by training ultrasonic video, which converts, includes:
For each video image in the trained ultrasonic video, building and the video image same size with
Machine image；
To the random image carry out convolution, and according to after convolution image and the deformation intensity factor obtain displacement diagram
Picture；
The displacement image is acted on the video image, the transformed video image of described image is obtained.
It is described in detail for realizing image transformation for the flexible deformation under ultrasonic contrast mode below.
Specifically, the step of flexible deformation includes:
The radiography part in original image is only cut out, the image of two-dimensional portion does not use.Random file is created first
Make anamorphose, establish and original image image of a size, picture traverse w is highly h, each pixel seat
Establish the random field of (- 1 ,+1) Δ x (x, y)=rand at mark, rand indicates random number functions, indicate to generate herein -1~1 it
Between random number, by the value assignment of the random number in the pixel value of the coordinate, then using standard deviation be σ Gaussian function to this
The image of generation carries out convolution, by the scale factor by the result of Gaussian convolution multiplied by control deformation intensity, obtains a bullet
The displacement field of property deformation.Finally on the original image by the effect of this displacement field, the data of final elastic deformation enhancing are obtained, and
And the coordinate of mark can also obtain transformed groundtruth (true value) according to this transformation.Such as Fig. 9 and Figure 10 institute
Show, wherein Fig. 9 indicates that the image before image transformation, Figure 10 indicate the transformed image of image.
It should be noted that the Feature Selection Model in the present embodiment is specifically as follows network model (for example, convolution
Neural network), image scaling to 299*299 resolution ratio is carried out in advance as the input of convolutional neural networks model with the network
It surveys.The present invention need to only obtain the network output of layer second from the bottom, and output dimension is 2048, by PCA (Principal
Component Analysis, principal component analysis) it is 1024 dimensions after dimensionality reduction, and save into file.The present embodiment is with ultrasonic contrast
It is respectively acquired 100 frames (10s) with the training ultrasonic video of capilary imaging technique, two-dimensional ultrasound, colorful blood and elastogram
For training ultrasound image respectively acquires 3 frames, it is as follows that single sample calculates available corresponding characteristic dimension:
Ultrasonic contrast (CEUS): 100 × 1024；
Capilary imaging technique: 100 × 1024；
Elastogram (UE): 3 × 1024；
Two-dimensional ultrasound (B-MODE): 3 × 1024；
C mode namely colorful blood (C-MODE): 3 × 1024.
As another specific embodiment of the present embodiment, the second initialization network includes characteristic aggregation model and mixes
Disaggregated model is closed,
The characteristic aggregation model is used to export the characteristic value after polymerization；
The hybrid classification model is used for the property class according to the characteristic value output patch after polymerization.
Specifically, as shown in figure 11, the above-mentioned multiple mode training ultrasonic videos of the second initialization network inputs and instruction
Practice the feature of 1024 dimensions of ultrasound image, after port number is hands-on ultrasonic video and/or trains ultrasound image down-sampled
Frame number, in the present embodiment, the port number of ultrasonic contrast and capilary imaging technique is 100, elastogram, two-dimensional ultrasound and coloured silk
Color blood flow is 3.The network model mainly includes two parts i.e. characteristic aggregation model and hybrid classification model.Characteristic aggregation model
The parameter list for establishing available network model learning acquires its cluster centre and associated weight with the method for deep learning automatically.
Input after characteristic aggregation is entered back into a special sorter network, final classification results are obtained.
It should be noted that the characteristic aggregation model compression concrete mode are as follows:
Wherein, the characteristics of image of N number of D dimension is inputted, V (j, k) indicates the value after final characteristic aggregation.N indicates that arteria carotis is super
The frame number of each mode video image of sound, dynamic video and still image frame number are respectively N1 and N2, x hereiDimension be D, D
It is 1024, is the result of above-mentioned network model convolutional neural networks output.ckIndicate cluster centre, akIndicate weight parameter, that is,
Final output vector V storage is residual sum of all x in its corresponding cluster, carries out networking to above-mentioned formula,
That is:
Wherein, wk、bkAnd ckIndicate the training parameter of network model, dimension is respectively D*K, K and D*K, and K is indicated in cluster
The number of the heart, K of the invention for radiography and this dynamic video of capilary imaging technique for 128, for still image
Elastogram, two-dimensional ultrasound and colorful blood for 32,The process for indicating SoftMax, by model
SoftMax this layer can realize.Specifically, the weight learning matrix m1 for initially setting up D*K dimension size, multiplied by input
The parameter x of model, in addition b obtains a by SoftMax network layerk, which is N*K, and wherein N indicates frame number, akMultiplied by
Input x obtains intermediate output p1, calculates akThe sum of penultimate dimension is multiplied by ckIntermediate output p2 is obtained, p1-p2 is the mould
The output V of type, the above process can train the w of place's optimum efficiency by the method for neural network iterationk、bkAnd ck。
After the dynamic and still image of each mode pass through characteristic aggregation network, the feature of output is spliced, is obtained
Length is the feature of (k1+k1+k2+k2+k2) * D size, and wherein k1 is 128, k2 32.
Into before hybrid classification model, then adds a hidden layer and further compressed for feature, above-mentioned output dimension
For 1* (k1+k1+k2+k2+k2) * D, the parameter of the hidden layer is that (k1+k1+k2+k2+k2) * D*D2, D2 are specific in the present invention
It can be 1024, obtaining final output dimension by matrix multiplication is 1*D2, i.e., the output of final entire ultrasonic contrast video is special
The size for levying vector is 1024.
Wherein, D indicates the dimension of characteristics of image, and k1 indicates the number of the cluster centre of dynamic video, indicates ultrasound herein
The number of radiography and the cluster centre under capilary imaging technique；K2 indicates the number of the cluster centre of still image, herein table
Show the number of the distance center under elastogram, two dimensional image and blood-stream image.
Mixed Weibull distribution is the network of a classification, is realized in network model by several fully-connected networks, specific as schemed
Shown in 12.
Method shown in Figure 12 specifically:
Wherein, X indicates the compressed output of features described above, dimension 1024, and Net1, Net2 and Net3 indicate three and connect entirely
Connect network, Net1 and Net2 are fully-connected networks, and Net1 and the fully-connected network of Net2 its parameter are respectively WkAnd ck, and only
One layer.Its parameter of this fully-connected network of Net3 isAlso there was only one layer, y indicates the output of final classification, dimension 2,
And encoded for one-hot, optionally, K value of the invention is 2.The network can be by way of two Net (Net1 and Net2)
Each self study then in such a way that a door network (Net3) is eventually by summation merges multiple models to the distinction of feature
Output is as a result, parameter above can be trained by way of neural network and be obtained.
It should be noted that according to n trained ultrasonic video and/or m trained ultrasound image and second mark
Information training the second initialization network may include: to set random between 0 to 1 for the weight parameter of the second initialization network
Number, training image use the data after convolutional neural networks model extraction, training iteration to loss minimum.
It should be noted that the equipment of the stability of above-mentioned differentiation patch and training first nerves network and nervus opticus net
The equipment of network may be same equipment, it is also possible to not be same equipment, the present embodiment is to this and without limitation.
As another embodiment of the present invention, a kind of storage medium is provided, wherein be stored at least in the storage medium
One program instruction, at least one program instruction is for being loaded and being executed by processor to realize that neck as previously described moves
The property method of discrimination of arteries and veins patch.
It should be understood that storage medium provided by the invention can be stored for executing previously described carotid plaques
Property method of discrimination program instruction, therefore can be provided for analysis and diagnosis of the medical staff to carotid plaques important
Foundation.
As another embodiment of the present invention, provide a kind of Vltrasonic device, wherein the Vltrasonic device include processor and
Memory, the memory include previously described storage medium, and the processor is for loading and executing the storage medium
On program instruction.
Vltrasonic device provided by the invention is stored due to using memory above for executing previously described neck
The program instruction of the property method of discrimination of artery plaque, therefore can effectively identify the property class of the blood vessel of patch, it is real
The qualitative analysis to the blood vessel of arteria carotis is showed, to provide for analysis and diagnosis of the medical staff to carotid plaques important
Foundation.
It is understood that the principle that embodiment of above is intended to be merely illustrative of the present and the exemplary implementation that uses
Mode, however the present invention is not limited thereto.For those skilled in the art, essence of the invention is not being departed from
In the case where mind and essence, various changes and modifications can be made therein, these variations and modifications are also considered as protection scope of the present invention.
Claims (10)
1. a kind of property method of discrimination of carotid plaques, which is characterized in that the property method of discrimination packet of the carotid plaques
It includes:
At least two groups ultrasound information that same clients include carotid plaques is obtained, every group of ultrasound information includes that clients are based on
Any one obtained ultrasonic video or ultrasound image in a variety of ultrasonic examination modes, and the ultrasound of at least two groups ultrasound information
Test mode is different；
By at least two groups ultrasound information described in first nerves Network Recognition, at least two groups vessel information, blood described in every group are obtained
Pipe information includes blood vessel video or blood-vessel image, and the blood vessel video includes the blood vessel region for having patch in ultrasonic video
Image stream, the blood-vessel image include the picture having where the blood vessel of patch in ultrasound image；
The characteristics of image in at least two groups vessel information is extracted by Feature Selection Model；
According to the characteristics of image in nervus opticus network and at least two groups vessel information, identification obtains the property of patch in blood vessel
Matter classification.
2. the property method of discrimination of carotid plaques according to claim 1, which is characterized in that the carotid plaques
Property method of discrimination further include:
Training sample is obtained, the training sample includes the n group training ultrasound information and every group of training ultrasound letter of m clients
The markup information of breath, and each clients correspond at least two groups training ultrasound information, every group of trained ultrasound information includes that training is super
Sound video or training ultrasound image, the markup information include the first markup information and the second markup information, first mark
Information is for marking the position having where the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image, and described the
Two markup informations are used to mark the property class of patch in the trained ultrasonic video or the trained ultrasound image, n be greater than
Integer equal to 2, m are the integer more than or equal to 1；
According to n group training ultrasound information and first markup information training the first initialization network, described the is obtained
One neural network；
According to n group training ultrasound information and second markup information training the second initialization network, described the is obtained
Two neural networks.
3. the property method of discrimination of carotid plaques according to claim 1, which is characterized in that the ultrasonic examination mode
For two-dimensional ultrasound, capilary imaging technique, colorful blood, elastogram or ultrasonic contrast.
4. a kind of property method of discrimination of carotid plaques, which is characterized in that the property method of discrimination packet of the carotid plaques
It includes:
Training sample is obtained, the training sample includes the n group training ultrasound information and every group of training ultrasound letter of m clients
The markup information of breath, every group of trained ultrasound information includes training ultrasonic video or training ultrasound image, and each clients are corresponding
At least two groups train ultrasound information, and the markup information includes the first markup information and the second markup information, first mark
Information is for marking the position having where the blood vessel of patch in the trained ultrasonic video or the trained ultrasound image, and described the
Two markup informations are used to mark the property class of patch in the trained ultrasonic video or the trained ultrasound image, n be greater than
Integer equal to 2, m are the integer more than or equal to 1；
According to n group training ultrasound information and first markup information training the first initialization network, the first mind is obtained
Through network, the first nerves network has the blood of patch in the trained ultrasonic video or the trained ultrasound image for identification
Position where pipe；
The characteristics of image in the n group training ultrasound information is extracted, characteristic aggregation is carried out to the characteristics of image extracted, according to spy
Characteristics of image and second markup information training the second initialization network after sign polymerization, obtain nervus opticus network, institute
State the property class of nervus opticus network patch for identification.
5. the property method of discrimination of carotid plaques according to claim 4, which is characterized in that if n group ultrasound instruction
Practicing includes the ultrasonic training information for meeting preset condition in information, described to be marked according to n group training ultrasound information and described first
Infuse information training the first initialization network, comprising:
By in the ultrasonic training information for meeting the preset condition two-dimensional image information and other types image information melt
It closes；
According to fused information and first markup information training the first initialization network.
6. the property method of discrimination of carotid plaques according to claim 5, which is characterized in that it is described will meet it is described pre-
If the two-dimensional image information and other types image information in the ultrasonic training information of condition are merged, comprising:
After the gray value of position in the two-dimensional image information and other types image information is determined as fusion respectively
Image corresponding position channel B value；
The gray value of position in the two-dimensional image information is determined as fused image in the channel G of corresponding position
Value；
Default value is determined as value of the fused image at the channel R.
7. the property method of discrimination of carotid plaques according to claim 4, which is characterized in that the described pair of figure extracted
As feature progress characteristic aggregation, according to the second initialization of characteristics of image and second markup information training after characteristic aggregation
Network obtains nervus opticus network, comprising:
The characteristics of image extracted is subjected to merging features, obtains spliced characteristics of image；
According to the spliced characteristics of image and second markup information training the second initialization network, the second mind is obtained
Through network.
8. according to the property method of discrimination of the carotid plaques as claimed in claim 7, which is characterized in that extract the n group
Characteristics of image in training ultrasound information carries out characteristic aggregation to the characteristics of image extracted, according to the image after characteristic aggregation
Feature and second markup information training the second initialization network, before obtaining nervus opticus network, further includes:
Video image in the trained ultrasonic video is carried out image to convert to obtain the transformed video image of described image,
Described in image transformation include translation, overturning, rotation and at least one of flexible deformation；
The characteristics of image in the transformed video image of described image is extracted by Feature Selection Model.
9. a kind of storage medium, which is characterized in that it is stored at least one program instruction in the storage medium, described at least one
Bar program instruction is used to load and executed by processor the arteria carotis spot with realization as described in any one of claims 1 to 3
The property method of discrimination of block, alternatively, realizing that the property of the carotid plaques as described in any one of claim 4 to 8 differentiates
Method.
10. a kind of Vltrasonic device, which is characterized in that the Vltrasonic device includes processor and memory, and the memory includes
Storage medium as claimed in claim 9, the processor is for loading and executing the program instruction on the storage medium.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910744483.9A CN110428417A (en) | 2019-08-13 | 2019-08-13 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910744483.9A CN110428417A (en) | 2019-08-13 | 2019-08-13 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110428417A true CN110428417A (en) | 2019-11-08 |
Family
ID=68414312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910744483.9A Pending CN110428417A (en) | 2019-08-13 | 2019-08-13 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110428417A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838110A (en) * | 2019-11-05 | 2020-02-25 | 张峰 | System for identifying benign and malignant tumor based on ultrasonic imaging |
| CN112215836A (en) * | 2020-10-22 | 2021-01-12 | 深圳市第二人民医院(深圳市转化医学研究院) | Carotid plaque detection method and device based on medical ultrasonic image |
| CN112365474A (en) * | 2020-11-12 | 2021-02-12 | 中科麦迪人工智能研究院(苏州)有限公司 | Blood vessel extraction method, device, electronic equipment and storage medium |
| CN113749690A (en) * | 2021-09-24 | 2021-12-07 | 无锡祥生医疗科技股份有限公司 | Blood flow measuring method and device for blood vessel and storage medium |
| CN114092744A (en) * | 2021-11-26 | 2022-02-25 | 山东大学 | Carotid artery ultrasound image plaque classification detection method and system |
| CN114332040A (en) * | 2021-12-30 | 2022-04-12 | 华中科技大学协和深圳医院 | Multi-mode-based thyroid tumor image classification method and terminal equipment |
| CN114469174A (en) * | 2021-12-17 | 2022-05-13 | 上海深至信息科技有限公司 | Artery plaque identification method and system based on ultrasonic scanning video |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1929781A (en) * | 2003-08-21 | 2007-03-14 | 依斯克姆公司 | Automated methods and systems for vascular plaque detection and analysis |
| CN102548481A (en) * | 2009-09-30 | 2012-07-04 | 富士胶片株式会社 | Diagnosis assisting system, diagnosis assisting program, and diagnosis assisting method |
| US20160110867A1 (en) * | 2010-08-12 | 2016-04-21 | Heartflow, Inc. | Method and system for image processing to determine patient-specific blood flow characteristics |
| CN106056126A (en) * | 2015-02-13 | 2016-10-26 | 西门子公司 | Plaque vulnerability assessment in medical imaging |
| 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 |
| CN109002838A (en) * | 2018-06-21 | 2018-12-14 | 山东大学 | IVUS vulnerable plaque tissue classification procedure and system based on vulnerability index |
| CN109285158A (en) * | 2018-07-24 | 2019-01-29 | 深圳先进技术研究院 | Vascular wall patch dividing method, device and computer readable storage medium |
-
2019
- 2019-08-13 CN CN201910744483.9A patent/CN110428417A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1929781A (en) * | 2003-08-21 | 2007-03-14 | 依斯克姆公司 | Automated methods and systems for vascular plaque detection and analysis |
| CN102548481A (en) * | 2009-09-30 | 2012-07-04 | 富士胶片株式会社 | Diagnosis assisting system, diagnosis assisting program, and diagnosis assisting method |
| US20160110867A1 (en) * | 2010-08-12 | 2016-04-21 | Heartflow, Inc. | Method and system for image processing to determine patient-specific blood flow characteristics |
| CN106056126A (en) * | 2015-02-13 | 2016-10-26 | 西门子公司 | Plaque vulnerability assessment in medical imaging |
| 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 |
| CN109002838A (en) * | 2018-06-21 | 2018-12-14 | 山东大学 | IVUS vulnerable plaque tissue classification procedure and system based on vulnerability index |
| CN109285158A (en) * | 2018-07-24 | 2019-01-29 | 深圳先进技术研究院 | Vascular wall patch dividing method, device and computer readable storage medium |
Non-Patent Citations (4)
| Title |
|---|
| KOLLURU, C., ET.AL: "Deep neural networks for A-line-based plaque classification in coronary intravascular optical coherence tomography images", 《JOURNAL OF MEDICAL IMAGING》 * |
| 吴宇鹏: "基于深度学习网络的血管内超声影像斑块识别与评估方法研究", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
| 孙夏，等: "基于卷积神经网络的颈动脉斑块超声图像特征识别", 《中国医疗器械信息》 * |
| 杨鑫: "三维超声图像中颈动脉粥样硬化的表型量化与分析", 《中国优秀博士学位论文全文数据库》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838110A (en) * | 2019-11-05 | 2020-02-25 | 张峰 | System for identifying benign and malignant tumor based on ultrasonic imaging |
| CN112215836A (en) * | 2020-10-22 | 2021-01-12 | 深圳市第二人民医院(深圳市转化医学研究院) | Carotid plaque detection method and device based on medical ultrasonic image |
| CN112365474A (en) * | 2020-11-12 | 2021-02-12 | 中科麦迪人工智能研究院(苏州)有限公司 | Blood vessel extraction method, device, electronic equipment and storage medium |
| CN113749690A (en) * | 2021-09-24 | 2021-12-07 | 无锡祥生医疗科技股份有限公司 | Blood flow measuring method and device for blood vessel and storage medium |
| CN113749690B (en) * | 2021-09-24 | 2024-01-30 | 无锡祥生医疗科技股份有限公司 | Blood vessel blood flow measuring method, device and storage medium |
| CN114092744A (en) * | 2021-11-26 | 2022-02-25 | 山东大学 | Carotid artery ultrasound image plaque classification detection method and system |
| CN114092744B (en) * | 2021-11-26 | 2024-05-17 | 山东大学 | Carotid ultrasonic image plaque classification detection method and system |
| CN114469174A (en) * | 2021-12-17 | 2022-05-13 | 上海深至信息科技有限公司 | Artery plaque identification method and system based on ultrasonic scanning video |
| CN114469174B (en) * | 2021-12-17 | 2024-10-18 | 上海深至信息科技有限公司 | Arterial plaque identification method and system based on ultrasonic scanning video |
| CN114332040A (en) * | 2021-12-30 | 2022-04-12 | 华中科技大学协和深圳医院 | Multi-mode-based thyroid tumor image classification method and terminal equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110428417A (en) | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques | |
| US11101033B2 (en) | Medical image aided diagnosis method and system combining image recognition and report editing | |
| CN110310271A (en) | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques | |
| CN111227864B (en) | Device for detecting focus by using ultrasonic image and computer vision | |
| JP2022123076A (en) | Automated image acquisition for assisting user to operate ultrasound device | |
| Smistad et al. | Highlighting nerves and blood vessels for ultrasound-guided axillary nerve block procedures using neural networks | |
| CN100470560C (en) | A method for encoding image pixels, a method for processing images and a method for processing images aimed at qualitative recognition of the object reproduced by one or more image pixels | |
| JP5226978B2 (en) | Ultrasonic diagnostic apparatus and image processing program | |
| CN111214255B (en) | Medical ultrasonic image computer-aided method | |
| CN110338844A (en) | The display processing method and 3-D supersonic imaging method and system of three-dimensional imaging data | |
| CN104799882A (en) | Method and ultrasound apparatus for displaying ultrasound image corresponding to region of interest | |
| KR20090029673A (en) | Automated detection of planes from three-dimensional echocardiographic data | |
| Kok et al. | Articulated planar reformation for change visualization in small animal imaging | |
| CN102056547A (en) | Medical image processing device and method for processing medical image | |
| CN106798571A (en) | The method and ultrasonic device of ultrasonoscopy are provided | |
| US20190388057A1 (en) | System and method to guide the positioning of a physiological sensor | |
| Chiang et al. | A touchless interaction interface for observing medical imaging | |
| CN113274051B (en) | Ultrasonic auxiliary scanning method and device, electronic equipment and storage medium | |
| EP3689247A1 (en) | Ultrasound imaging method and ultrasound imaging system for carrying out the said method | |
| WO2024126468A1 (en) | Echocardiogram classification with machine learning | |
| Zhou et al. | Wrist ultrasound segmentation by deep learning | |
| KR101534088B1 (en) | Method for displaying ultrasound image using doppler data and ultrasound medical apparatus thereto | |
| US20230137369A1 (en) | Aiding a user to perform a medical ultrasound examination | |
| Srinivasan et al. | A comprehensive diagnostic tool for skin cancer using a multifaceted computer vision approach | |
| JP2002163635A (en) | System and method for supporting diagnosis of pervasive hepatic disease by utilizing hierarchical neural network on basis of feature amount provided from ultrasonic image of diagnostic part |
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: 20191108 |
|
| RJ01 | Rejection of invention patent application after publication |